Notesale: Turn your study into money

Document Preview

Extracts from the notes are below, to see the PDF you'll receive please use the links above

---

An Introduction to
Marketing Research

Scott M
...
Albaum
Emeritus Professor of Marketing
University of New Mexico

Copyright © 2010 by Scott M
...
Albaum
...

All rights reserved
...


Table of Contents
Chapter 1

An Introduction to Marketing Research
...
21

Chapter 3

Secondary Sources of Information
...
59

Chapter 5

Modes of Interviewing Personal-Send-Call
...
97

Chapter 7

Sampling Procedures in Research
...
153

Chapter 9

Measuring Respondent Information:
Attitudes, Satisfaction, Loyalty and Behavior
...
219

Chapter 11

Hypothesis Testing and Univariate Analysis
...
287

Chapter 13

Multivariate Statistical Analysis I
...
363

Chapter 15

Preparing the Research Report
...
G-1

Appendix 1
...

Green, Donald S
...
But, it is not a revision of that book
...
com
...
First, with respect to method, the
unifying concept of this book is that marketing research is a cost-incurring activity whose output
is information of potential value to managers in making decisions
...
The
entire book focuses on implementation of online marketing research
...
com, so that students can
design, plan and implement an online survey of their own at no charge
...
In terms of number of
chapters, 30% of the book is devoted to analysis, but, the discussion is at a level that senior-level
undergraduates can understand, and the techniques are explained within the context of computerbased analysis
...
This means that all
the basic elements of method, techniques, and analysis are covered, including those at a more
sophisticated level
...
The methodological scope
regarding research design, data collection techniques, and measurement is broad
...
The book presents its
material from primarily a pragmatic and user-oriented (rather than theoretical research technician)
perspective
...

Because the book is available online, it can be used in a modular fashion at no cost to the student
...
Similarly, if the course focuses on survey
research, chapters 9 and 10 could be the focal point, supplemented with chapters 1,2,4,5,6,7 plus
analysis chapters as appropriate
...

There is a Glossary of Terms and an appendix that includes some widely-used statistical tables
for analysis
...

Many people helped shape the content and style of this book, but most importantly Professors
Paul E
...
Tull have had a profound influence on the authors’ thinking
about research and their book with one of the present authors provided a platform from which the
present book was launched
...
SMITH
GERALD S
...
The role of marketing itself has
changed dramatically due to various crises—material and energy shortages, inflation, economic
recessions, high unemployment, dying industries, dying companies, terrorism and war, and
effects due to rapid technological changes in certain industries
...
The
means to help them do this is marketing research
...
It can be considered to be either primarily
fundamental or applied in nature
...
In contrast, applied
research, also known as decisional research, attempts to use existing knowledge to aid in the
solution of some given problem or set of problems
...
A
marketing manager must prioritize the more important and pressing problems selected for
solution, reach the best possible solution based on the information available, implement the
solution, modify the solution when additional information so dictates, and establish policy to act
as a ready-made solution for any recurrence of the problem
...

Figure 1
...
Smith and Gerald S
...
com Survey University
...
Each focuses on a different aspect of the “Company” and it’s interaction
with the “Customer” and “Competition” in the market environment:
Exhibit 1
...
Such studies provide key information about
market growth, competitive positioning and tracking share of market
...
This is often a
descriptive market segmentation and market share analysis
3 - Stage in the Purchase Process / Tracking Surveys
Where is the customer in the adoption process? This information shows market Awareness – Knowledge –
Intention – Trial – Purchase – Repurchase of the product
...
What motivates the customer to move from interest in the
product to actual purchase? This is a key to understanding customer conversion, commitment and loyalty
...
Used to direct advertising and improve customer conversion, commitment and loyalty
...

7 - New Product Concept Analysis Surveys
Concept test studies are appropriate in the initial screening of new product concepts
...

8 - New Product Acceptance and Demand Surveys (Conjoint Analysis)
Primarily for estimating demand for new products that can be described or have been developed in drawing or
concept, but have not yet been developed physically
...

9 - Habits and Uses Surveys
Directed at understanding usage situations, including how, when and where the product is used
...

10 - Product Fulfillment Surveys (Attribute, Features, Promised Benefits)
Evaluation of the product’s promised bundle of benefits (both tangible and image)
...


2
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

12 - Brand Equity Analysis Surveys
What is psychological value that a brand holds in the market place? Brand equity is a composite of brand
awareness, brand quality, brand associations and brand loyalty measures
...
Means-end analysis is often part of this type of study
...

15 - Sales Force Effectiveness Surveys
A combination of measures that focus on the sales activities, performance and effectiveness in producing the
desired and measurable effect or goal
...

16 - Sales Lead Generation Surveys
Sales lead generation surveys for (1) assuring timely use and follow-up of sales leads, (2) qualifying sales leads
(thereby saving valuable sales force time) and (3) providing more effective tracking of sales leads
...

18 - Customer Service Representative (CSR) Surveys: Attitudes, Burnout, Turnover and Retention:
CSRs hold attitudes that reflect on their job related activities including (1) the allocation of time; (2) solutions to
customer needs; (3) how to improve their job; (4) best practices; (5) How well internal departments help
customers
...

19 - Sales Forecasting and Market Tracking Surveys
Sales forecasting and market tracking studies can include expert opinion (experts estimate the market),
judgmental bootstrapping (expert based rules describing how to use available secondary market information),
conjoint analysis (estimation of consumer intentions based on product attributes that are important in the
decision), and intentions evaluations (consumer self reported intentions of future purchases) are to be made
...
Price surveys may estimate the demand for different product or service segments, or different usage
situations
...
qualtrics
...
php/Market_Surveys
...
This level of specificity is desirable for several reasons
...
Clarification
...

The need for more rigorous definitions of key variables often becomes apparent
...
Objectivity
...

3
...
Discussion helps to identify problems and common points of reference
when different people hold alternative implicit models of the same problem situation
...
Improvement of models
...
The degree and range of adaptability can thus be extended
...
Smith and Gerald S
...
Guide to research needs
...


While varying information is required for the different types of marketing research
projects, the key to conducting a successful research project lies with the researcher and the
client
...
This requires identifying the
appropriate questions, respondents, methodology, analysis and reporting
...
2)
...
2 Basic Research Issues
As technology advances, marketing researchers are continually looking for ways to adapt new technology
to the practice of research
...
However,
researchers must never forget that research basics cannot be overlooked
...
All studies must address the following basic
issues (Anderson, Berdie, & Liestman, 1984):
1
...
This is the essence of project design, and the heart of proper planning
...

2
...
Sample design should be such that only those people who are of interest to the
research user are contacted, and such that those who are contacted are reasonably representative of
the group of interest
...
Ask questions the right way
...
This is the essence of questionnaire design
...
One basic that is overlooked all too often is pretesting the questionnaire; this is
crucial for ensuring that responses are the ones that are needed to address the problem
...
Obtain answers to questions
...

Techniques used should be selected for how each bears on nonresponse and response alike
...
Relate answers to the needs of the research user/client
...

Proper data analysis is needed if a study is to have any value to the user
...

Common sense is a valuable tool for the researcher when considering alternative analysis approaches
for any project
...
Communicate effectively and in a way that the client understands
...
The information that is reported to the user should be in a form that is understandable to
the user so that he or she can tell that it is relevant to the issues at hand
...


4
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

THE BASIC RESEARCH PROCESS
How is marketing research actually conducted? What are the general steps in completing
a research project? These questions are answered in the steps of the research process
...
There are other times when “later”
decisions influence decisions that are made early in the research planning process
...
g
...

Figure 1
...
The researcher should write a formal plan for the project,
including the background information and statement of objectives, which then becomes the
master guide for implementing and controlling the research project
...


STAGE 1: PROBLEM FORMULATION
In a very real sense, problem formulation is the heart of the research process
...
From the researcher’s point of view,
problem formulation means translating the management problem into a research problem
...
This involves timely and clear
communication between manager and researcher
...
An accurate
problem formulation specifies the types of information needed to help solve the management
problem
...


5
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Exhibit 1
...
While hypotheses are crucial for basic research because they tell
the researcher what to do, the concept of a hypothesis can also be useful in decisional research to
direct the development of the research problem statement
...
Kerlinger and Lee (2000, Chapter
2) suggest that research problems and hypotheses meet the following criteria:
1
...

2
...

3
...

Where properties of good hypotheses include the following:
1
...

2
...
e
...

How to Formulate the Research Problem
Problem formulation is much easier when specific components of the research problem
are defined:
1
...
Objectives range from the very general, such as profit
maximization, to the highly specific, such as measuring market interest in a new product
...
However, the researcher needs to
take the initiative to develop a clear statement of objectives
...
Two or three well targeted objectives is preferable to many that are illconceived
...
If there are too
many objectives separate studies may be appropriate
...
Smith and Gerald S
...
The Environment or Context of the Problem
Consider the problem of deciding whether to introduce a new consumer product
...

The researcher’s efforts should be oriented toward helping the manager decide whether
any investigation is justified based on the potential value of the research findings versus their
cost
...

If the research is undertaken and if the resulting findings are to be utilized (i
...
, have an
influence on the user’s decision making), the manager and researcher must have a productive and
trusting relationship that is based on the researcher’s ability to perform and deliver the research
as promised
...
The Nature of the Problem
Every research problem may be evaluated on a scale that ranges from very simple to very
complex
...
Understanding the nature of the problem helps a researcher ensure that the right
problem is being investigated and that a marketing plan can be developed to solve the problem
...

4
...
Such a program of
action becomes a commitment, made in the present, to follow some behavioral pattern in the
future
...
A
hypothesis often implies a possible course of action with a prediction of the outcome if that
course of action is followed
...
This involves:
1
...
Determining the degree to which each variable can be controlled
3
...

The following example shows the results of a failure to follow through with these aspects
of the problem situation model
...
Smith and Gerald S
...
3 “New Coke” Versus Original Coke
In the mid-1980s the Coca Cola Company made a decision to introduce a new beverage product (Hartley, 1995,
pp
...
The company had evidence that taste was the single most important cause of Coke’s decline in the market
share in the late 1970s and early 1980s
...

Almost 200,000 blind product taste tests were conducted in the United States, and more than one-half of the
participants favored New Coke over both the original formula and Pepsi
...
This turned out to be a big mistake! Eventually, the company reintroduced the
original formula as Coke Classic and tried to market the two products
...

What went wrong? Two things stand out
...
Consumers were not told
that only one product would be marketed
...
Second, no one realized the symbolic value and emotional involvement people had with the original Coke
...

CBS New Coke News Clip: http://www
...
com/watch?v=-doEpVWFLsE&NR=1&feature=fvwp
New Coke Commercial: http://www
...
com/watch?v=o4YvmN1hvNA
New Coke and Coke Classic Commercial: http://www
...
com/watch?v=ky45YGUA3co

5
...
One of the manager’s primary jobs is to anticipate and
communicate the possible outcomes of various courses of action that may result from following
the research
...
Degrees of Uncertainty
Most marketing problems are characterized by a situation of uncertainty as to which course of
action is best
...

A carefully formulated problem and statement of research purpose is necessary for competent
research
...
The
research problem provides relevant information concerning recognized (or newly generated)
alternative solutions to aid in this choice
...
Even though this method is not the only one used, it is the standard against which other
investigative methods are measured
...
e
...
Its most distinctive characteristic is its total lack of
subjectivity
...
It is analytical in its processes and is investigatorindependent
...

8
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The goal of a scientific methodologist, also called an objectivist, is to run a hypothesis
test using publicly stated procedures that are investigator-independent
...
Although the steps are the same, there are
differences in the way in which the steps are performed and in the underlying assumptions about
behavior
...

This objectivist-subjectivist distinction has very practical meaning, particularly when
considering the use of outside research suppliers
...
Objectivist-based research is often called
quantitative research, whereas subjectivist-based research is often called qualitative research
...
4 The Scientific Method
In structure, if not always in application, the scientific method is simple and consists of the following steps:
1
...
This is the problem-awareness phase, which involves observing a set of significant
factors that relate to the problem situation
...
Formulation of hypotheses
...
e
...
In a sense, the
hypothesis suggests an explanation of what has been observed
...
Prediction of the future
...
This
stage uses the hypotheses to predict what will happen
...
Testing the hypotheses
...
From a research project
perspective this is the design and implementation of the main study
...

A simple example will show how the scientific method works
...
Observation: The researcher notices some competitors’ sales are increasing and that many
competitors have shifted to a new plastic wrapping
...
Formulation of hypotheses: The researcher assumes his client’s products are of similar quality and
that the plastic wrapping is the sole cause of increased competitors’ sales
...
Prediction of the future: The hypothesis predicts that sales will increase if the manufacturer shifts to
the new wrapping
...
Testing the hypotheses: The client produces some shirts in the new packaging and market-tests
them
...
Smith and Gerald S
...
In
addition to selecting a method of inquiry, the research planner must also select a research
method
...
The major advantage of experimental research lies in
the ability to control extraneous variables and manipulate one or more variables by the
intervention of the investigator
...


STAGE 4: RESEARCH DESIGN
Research design is defined as the specific methods and procedures for acquiring the
information needed
...
Research designs are unique to a methodology
...


STAGE 5: DATA COLLECTION TECHNIQUES
Research design begins to take on detailed focus as the researcher selects the particular
techniques to be used in solving the problem formulated and in carrying out the method selected
...
Some techniques are unique to
a method of inquiry
...
In general, data collection uses either
communication or observation
...
This process can be
done in person, by mail, by telephone, by e-mail, and over the Internet
...
In contrast to this process, data
may be obtained by observing present or past behavior
...

In order to collect data from communication or observation there must be a means of
recording responses or behavior
...
The relationship is two-way
...


STAGE 6: SAMPLE DESIGN
Rarely will a marketing research project involve examining the entire population that is
relevant to the problem
...
g
...
value, etc
...

In other instances the use of a sample is derived from consideration of the relevant systematic
and variable errors that might arise in a project
...
Smith and Gerald S
...
Where the sample is to be selected
2
...
The size of the sample

The sample design must be consistent with the relevant population, which is usually
specified in the problem-formulation stage of the research process
...

The process of sample selection may be done by probability or non-probability methods
...
In contrast, a non-probability sample is one
selected on the basis of the judgment of the investigator, convenience, or by some other means
not involving the use of probabilities
...

Data collection, whether by communication or observation, requires the use of data collection
personnel which then raises questions regarding managing these people
...
Also, project design may require specialized data collection, which might best be
obtained from an outside supplier
...

A study of marketing research firms found that the major barriers to the communication
of information from clients to research suppliers to field service firms were insufficient
information supplied by the client, the research supplier as an intermediary between client and
field service firm, and lack of client interest in data collection (Segal & Newberry, 1983)
...
Another way to overcome communication
barriers is for the field service to be consulted on such major issues as scheduling, costs, and
purpose of the study
...
Although this study was conducted more than 20 years ago, these
are enduring problems that exist today
...
Data must be analyzed
...
The types of analyses that can be properly
performed depend upon the sampling procedures, measurement instruments, and data collection
techniques used
...


11
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

STAGE 9: THE RESEARCH REPORT
The culmination of the research process is the research report
...
Two critical attributes of the
report are that it provides all the information readers need using language they understand
(completeness) and that it contains selective information chosen by the researcher (conciseness)
...

Two approaches can be taken to ensure that this conflict is not a problem
...

The second approach is concerned with how the report is communicated
...
Thus, a written report, by itself, may
be inadequate and only an invitation to inaction
...
There are still more that, although they may
respond, will often misunderstand the meaning of what is written
...


RESOURCE PLANNING FOR YOUR STUDY
When planning for your research, the resources necessary to complete the study should
also be identified
...
Resource plans range from very
informal to very formal and may include a list all personnel who will be involved with the project, the
exact assignment of each person, the time to be spent, and the pay for each
...
Microsoft
Project or similar software may be helpful in planning and monitoring your research project
...
microsoft
...
aspx

12
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

THE MAKE OR BUY DECISION
A decision facing all companies that want to use marketing research is who should do the
research
...
In short, sourcing marketing research is a “make or buy” decision
...
Other companies with in-house capabilities treat the
internal units the same as outside suppliers: they must compete with outside suppliers by
preparing proposals and making bids for the business
...

Outside suppliers range from a full-service marketing research agency such as M/A/R/C
Research (http://www
...
com), Burke (http://www
...
com), and Maritz Marketing
Research, Inc
...
maritz
...
qualtrics
...
Clients can negotiate with full-service companies to perform
only limited services, for example, research design and data collection only, if that is all the
client wants
...

When might the use of an outside research supplier be appropriate? There are a number
of situations that may call for the use of such firms:
1
...

2
...

3
...

4
...

5
...

6
...
The outside research supplier can do the research quicker
...
Some aspect such as data collection may be cheaper when done by an outside supplier
(purchasing supermarket-based scanner data may be less costly than collecting it yourself)
...
There is a need for anonymity or confidentiality that may be provided best by an outside
research firm
...
The results of the research may be used in legal proceedings
...

Exhibit 1
...
For the most part these represent adaptation of more standard techniques and methods to fit a
B2B situation
...

2
...
This applies to all stages of a project
...

Lay the groundwork
...
In B2B situations there may be many people who have customer contact, and their views
may differ enough that it is beneficial to talk with as many as possible
...
Smith and Gerald S
...


Select and draw the samples
...
In a typical project it is often difficult to decide
which type of respondent to contact
...

4
...
Business respondents tend to be busy people, so it is important
to be creative in selecting data collection methodologies
...
Understanding how your target respondent
can best be contacted can be helpful in selecting the best method for the majority of the sample
...
Design the questions
...
This, of course, applies to all
research projects
...

6
...
Much of the information collected in B2B research is qualitative, making the
analysis crucial
...
Report the results
...
This includes use of individual respondent anecdotes and other humanizing details
...
Plan, communicate, and act
...
Communicating with customers allows a company
to involve them in implementing whatever action the research suggests
...


EXHIBIT 1
...
Both hardware and software are involved in such adaptations
...
Rather, what must be done is to adapt the new techniques and
technologies to these basics
...

2
...


4
...


6
...
This is the essence of project design, and the heart of proper planning
...

Ask the right people
...

Ask questions the right way
...
This is the essence of questionnaire design
...

Always pretest the questionnaire to ensure that responses are the ones that are needed to address the problem
...
The process of data collection is central to all marketing research
...

Relate answers to the needs of the research user/client
...
Proper data analysis
is needed if a study is to have any value to the user
...
Common sense is a valuable tool for the researcher when
considering alternative analysis approaches for any project
...
Many good projects are ruined in this stage
...


14
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Ethical Considerations in Survey Research
There are a number of ethical considerations that arise both in conducting marketing
research projects and in marketing related activities in general
...
Exhibit 1
...

EXHIBIT 1
...
An additional concern too frequent interviewing of the respondent
...
e
...
Respondents have rights that should not be ignored or violated
...
The researcher is often in a position of authority and as such should
assure that the participant does not feel forced to comply, has the ability to choose and make
informed choices, is safe from stress, psychological and physical harm, providing information
detrimental to their self-interest, and have the right to be informed of the purpose of the research
...
Smith and Gerald S
...
Exhibit 1
...

The American Marketing Association has provided a statement identifying principles of
ethical practice of marketing research (Exhibit 1
...
These broad guidelines provide standards
for the protection of the marketer and respondent alike and even extend to researchers and to
marketers who are not engaged in research activities
...
8 Ethical Questions Regarding Subjects’ Rights
Subjects’ Rights
A
...
Awareness of right
b
...
Opportunity to make a choice

Possible Results of Violation of Rights
a
...

c
...


The right to be safe
a
...
Subjects’ right to be free from stress

a
...


C
...
Debriefing

a
...


Dissemination of data

b
...


Right to not be deceived

c
...
M
...
, in Journal of Marketing, 11, p
...
November, 1974
...
Smith and Gerald S
...
9 Ethics in Marketing Research

The American Marketing Association on
Ethics in Marketing Research
The American Marketing Association has established principles of ethical practice of marketing research for the
guidance of its members
...
Similarly, the research practitioner has an
obligation to the discipline and to those who provide support for it—an obligation to adhere to basic and
commonly accepted standards of scientific investigation as they apply to the domain of marketing research
...
No individual or organization will undertake any activity which is directly or indirectly represented to be
marketing research, but which has as its real purpose the attempted sales of merchandise or services to
some or all of the respondents interviewed in the course of the research
...
If respondents have been led to believe, directly or indirectly, that they are participating in a marketing
research survey and that their anonymity will be protected, their names shall not be made known to any
one outside the research organization or research department, or used for other than research purposes
...
There will be no intentional or deliberate misrepresentation of research methods or results
...

Evidence that fieldwork has been completed according to specifications will, upon request, be made
available to buyers of the research
...
The identity of the survey sponsor and/or the ultimate client for whom a survey is being done will be held in
confidence at all times, unless this identity is to be revealed as part of the research design
...

3
...

FOR USERS OF MARKETING RESEARCH
1
...

2
...

FOR FIELD INTERVIEWERS
1
...

2
...

3
...

4
...
Members of the American Marketing Association will be expected
to conduct themselves in accordance with the provisions of this code in all of their marketing research
activities
...
Smith and Gerald S
...
Planning a
research project includes:
Problem formulation
Method of inquiry
Research method
Research design
Selection of data collection techniques
Sample design
Data collection
Analysis and interpretation of data
Research reports

The differences in client requirements give rise to different requirements in research projects
...

In this chapter we explained how managers use marketing research to help decision-making
...
There are several groups who “do” marketing research but the
basic reason for their research is the same—to solve a problem
...
Information should be accurate,
current, sufficient, available and, most important, relevant to be meaningful to organizations
...

The dialogue can encompass objectives, courses of action, and environmental variables affecting
decision outcomes
...
Although there are many
concerns among researchers and others, they can be summarized as deceptive/fraudulent
practices, invasion of privacy, and lack of consideration for subjects/respondents
...
Professional codes of conduct of marketing research were presented, which
are indicative of an industry that is trying to “clean up its own act
...
Smith and Gerald S
...
marketingpower
...
aspx?sq=ethics+statement
Anderson, J
...
, Berdie, D
...
, & Liestman, R
...
Hi-tech techniques OK, but don’t forget research
basics
...
2), 12
...
youtube
...
(1966, March–June)
...
subjectivism in the social sciences
...

Kerlinger, F
...
, & Lee, H
...
Belmont, CA: Wadsworth
Publishing
...
microsoft
...
aspx
New Coke Commercial: http://www
...
com/watch?v=o4YvmN1hvNA
New Coke and Coke Classic Commercial: http://www
...
com/watch?v=ky45YGUA3co
Qualtrics
...
qualtrics
...
php/Market_Surveys
...
(2001, November 19)
...
Marketing News, 35, 17–
19
...
C
...
Subjects and respondent abuse in marketing research
...

13–19
...
N
...
(1983,Winter)
...
Journal of Data Collection, 23, 58–59
...
M
...
(1974)
...
359
...
Published by the American Marketing Association

19
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

This page is intentionally blank

20
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 2
DEFINING THE RESEARCH DESIGN AND
CONTROLLING RESEARCH ERRORS
A research design specifies the methods and procedures for acquiring the information
needed to structure and solve the research problem
...
A good research design ensures that the information obtained is relevant to the
research problem, and that it is collected by objective and economical procedures
...

Research designs vary depending on the type of study
...
Each will be described in turn
...

An exploratory study is often the first in a series of projects
...
Thus flexibility is a key to
designing and conducting exploratory studies
...
It is the rare research
problem for which there is no relevant information to be found by a relatively quick and
inexpensive search of the literature
...
Searches should also be made of company records
...
In focus group
interviews, a group of knowledgeable people participate in a joint interview that does not use a
structured question-and-answer methodology
...
Smith and Gerald S
...
The interviewer or moderator of the focus group session works with the client to
develop a general discussion outline that typically includes such topics as usage experience,
problems with use, and how decisions are made
...

Focus groups are used primarily to identify and define problems, provide background
information, and generate hypotheses
...
Areas of application include detecting trends in lifestyles, examining new product or
service concepts, generating ideas for improving established products or services, developing
creative concepts for advertising, and determining effective means of marketing products or
services
...
Limited research on this issue conducted more than 20 years ago
suggests that the number and quality of ideas generated may be greater from such interviews
(Fern, 1982)
...
Identifying and understanding consumer language relating to the product category in question
...
Identifying the range of consumer concerns
...
Identifying the complexity of consumer concerns
...
Identifying specific methodological or logistical problems that are likely to affect either the
cost of the subsequent research, or one’s ability to generate meaningful, actionable findings
...

Examine Analogous Situations

It is also logical that a researcher will want to examine analogous situations to determine
what else can be learned about the nature of the problem and its variables
...
More discussion of the use of focus groups is given in
Chapter 4
...
For
example, a market potential study may describe the number, distribution, and socioeconomic
characteristics of potential customers of a product
...
A sales analysis study describes
sales by territory, type of account, size or model of product, and the like
...
Smith and Gerald S
...
Descriptive
designs, often called observational designs by some researchers, provide information on groups
and phenomena that already exist; no new groups are created (Fink, 2003)
...
Management knew
very little about the members, other than that they were school employees, family members of
employees, or former employees
...
Consequently,
investigators undertook a study to answer the following research questions:
1
...
How extensively are existing services being used, and what are members’ attitudes toward such
services?
3
...
It is not always necessary to
understand causal relations in order to make accurate predictive statements
...
The basic principle involved is to find
desirable behavior correlates that are measurable when the predictive statement is made
...
” Further, we
would like to know the relationships of these causal factors to the effects that we are predicting
...

Bases for Inferring Causal Relationships

There are three types of evidence that can be used for drawing inferences about causal
relationships:
1
...
Sequence of events
3
...
Smith and Gerald S
...
That is, there must be logical
implication (or theoretical justification) to imply the specific causal relation
...
Two types of associative
variation may be distinguished:
1
...
Association by change: A measure of the extent to which a change in the level of one
variable is associated with a change in the level of the other
...

Sequence of Events

A second characteristic of a causal relationship is the requirement that the causal factor
occur first; the cause must precede the result
...

Absence of Other Possible Causal Factors

A final basis for inferring causation is the absence of any other possible causes other than
the one(s) being investigated
...

Obviously, in an after-the-fact examination of an uncontrolled result such as an increase
in detergent sales, it is impossible to clearly rule out all causal factors other than salesperson
retraining
...

Conclusions Concerning Types of Evidence

No single type of evidence, or even the combination of all three types considered, can
ever conclusively demonstrate that a causal relationship exists
...
However, we can obtain evidence that makes it highly reasonable to conclude that a
particular relationship exists
...
1 shows certain questions that are necessary to answer
...
Smith and Gerald S
...
1 Issues in Determining Causation
Several questions arise when determining whether a variable X has causal power over another variable, Y:
1
...
What is the direction of causality—does X positively influence Y, or is the relationship negative?
3
...
Which value of the believed cause exerts a causal influence—its presence or absence?
5
...
(Monroe and Petroshius, n
...


SOURCES OF MARKETING INFORMATION
There are five major sources for obtaining marketing information
...

Secondary sources
Respondents
Natural experiments
Controlled experiments
Simulation

Secondary Sources of Information
Secondary information is information that has been collected by persons or agencies for
purposes other than the solution of the problem at hand
...

The federal government collects and publishes information on the numbers of families, family
formation, income, and the number and sales volume of retail stores, all by geographic area
...
Many state and local governments collect
similar information for their respective areas
...
Trade journals are also a valuable source of secondary
information, as are special studies done by other advertising media
...
These so-called syndicated services, particularly those for packaged
consumer goods, are becoming more sophisticated as they are increasingly becoming based on
25
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

scanner data
...

Information from Respondents
A second major source of information is obtained from respondents
...

The term respondent literally means “one who responds or answers
...


In this book we shall consider both the information obtained from asking people
questions, and that provided by observing behavior (or the results of past behavior) to comprise
information from respondents
...
While both natural and controlled experimental designs are capable of
providing associative variation and sequence of events, only controlled experiments can provide
reasonably conclusive evidence concerning the third type of evidence, the absence of other
possible producers
...
That is, there is no manipulation of an assumed causal variable
...
As such, the natural experiment is a form of ex
post facto research
...
The variable of interest has occurred in a natural
setting, and the researcher looks for respondents who have been exposed to it and also, if a
control group is desired, respondents who have not been exposed
...
For example, if the
impact of a television commercial on attitudes were desired, the investigator would conduct a
survey of people after the commercial was shown
...

Differences in attitudes could be compared as a crude measure of impact
...
For a brief discussion of natural experiments, see
Anderson (1971)
...
Specifically, two kinds of intervention are required:
1
...
Random assignment of subjects to experimental and control groups
The researcher conducts the experiment by assigning the subjects to an experimental
group where the causal variable is manipulated, or to a control group where no manipulation of
the causal variable occurs
...
Differences between the groups, if present, are
attributed to the manipulation variable
...
Smith and Gerald S
...
For
example, researchers often use the advanced branching logic, randomization, question block
presentation, question timing, and java scripting capabilities of Qualtrics
...

Simulation
The expense and time involved in the personal interviews often associated with field
experimentation may preclude it as a source of information for a particular operational situation
...
The manipulation of such
models is called simulation
...
Models
that are environmentally rich (that is, that contain many complex interactions and nonlinear
relationships among the variables, probabilistic components, time dependencies, etc
...
Rather, the analyst views a simulation model as a limited imitation of
the process or system under study and attempts to run the system on a computer to see what
would happen if a particular policy were put into effect
...
During the past 50 or more years, simulations have been developed for
such marketing decision- making applications as marketing systems, marketing-mix elements
(new-product, price advertising, and sales-force decisions), and interviewing costs in marketing
surveys
...

Potential errors can arise at any point from problem formulation through report preparation, and
rarely will a research project be error-free
...
As we shall see in the next section of this
chapter, there are alternative strategies that can be followed
...
Maximizing accuracy requires that total study errors be minimized
...
Since this introduces random
variability into the precision with which a sample statistic is calculated, it is often called random
sampling error
...
2 gives an illustration of how total error is assessed
...
Smith and Gerald S
...
2 How Errors Add Up
It is important to know all the sources of error that contribute to inaccuracy, and to assess the
impact of each
...

When estimating the market, adjustments are made for each source of error
...
For purposes of simplicity, total error is
shown here as the sum of the component errors
...
Assessing the individual
components of total error is highly judgmental and subjective, but it is worth the effort
...
Such
errors are diverse in nature and are often thought of as resulting in some sort of bias, which
implies systematic error
...
However, there can
be a random component of non-sampling error
...
Non-sampling errors have both nonresponse and response based origins
...
Smith and Gerald S
...
Increasing the
sample size to reduce sampling error can increase non-sampling error in that, for example, there
are more instances where such things as recording errors can occur, and the impact of biased
(i
...
, nonobjective) questions and other systematic errors will be greater
...

Ideally, efforts should be made to minimize each component
...
The researcher must make a decision that involves a tradeoff
between sampling and non-sampling errors
...
In a study comparing several research designs and data
collection methods, Assael and Keon (1982) concluded that non-sampling error far outweighs
random sampling error in contributing to total survey error
...
3 briefly defines eight major types of errors that can influence
research results
...
3 Types of Errors in the Research Process
Different types of errors can influence research results:
Population specification: noncorrespondence of the required population to the population selected by
the researcher
Sampling: noncorrespondence of the sample selected by probability means and the representative
sample sought by the researcher
Selection: noncorrespondence of the sample selected by nonprobability means and the sought
representative sample
Frame: noncorrespondence of the sought sample to the required sample
Nonresponse: noncorrespondence of the achieved (or obtained) sample to the selected sample
Surrogate information: noncorrespondence of the information being sought by the researcher and that
required to solve the problem
Measurement: noncorrespondence of the information obtained by the measurement process and the
information sought by the researcher
Experimental: noncorrespondence of the true (or actual) impact of, and the impact attributed to, the
independent variable(s)

Population Specification Error
This type of error occurs when the researcher selects an inappropriate population or
universe from which to obtain data
...
They consider conducting a survey of purchasing agents from major
corporations presently owning such aircraft
...

Similarly, packaged goods manufacturers often conduct surveys of housewives, because they are
easier to contact, and it is assumed they decide what is to be purchased and also do the actual
purchasing
...
The husband may purchase
a significant share of the packaged goods, and have significant direct and indirect influence over what
is bought
...
Smith and Gerald S
...

Example: Suppose that we collected a random sample of 500 people from the general adult
population and upon analysis found it to be composed only of people aged 35 to 55
...
Sampling error is affected by the
homogeneity of the population being studied and sampled from and by the size of the sample
...
If a census were conducted (i
...
, all elements of the population were
included) there would be no sampling error
...

Example: Interviewers conducting a mall intercept study have a natural tendency to select those
respondents who are the most accessible and agreeable whenever there is latitude to do so
...


Selection error often reflects people, who are most easily reached, better dressed, and
have better kept homes or more pleasant personalities
...

Frame Error
A sampling frame is the source for sampling that represents all the members of the
population
...

Example: Consider the sample frame for a shopper intercept study at a shopping mall
...
A commonly used frame
for consumer research is the telephone directory
...


A perfect frame identifies each member of the population once, but only once, and does
not include members not in the population of interest
...
There are two ways in which nonresponse can occur: (a) noncontact (the inability to
contact all members of the sample); and (b) refusal (nonresponse to some or all items on the
measurement instrument)
...

30
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Example: In telephone surveys, some respondents are inaccessible because they are not at home
(NAH) for the initial call or call-backs
...
Not-at-home respondents are typically younger with no small children, and have a much
higher proportion of working wives than households with someone at home
...
Thus, most surveys can anticipate errors from non-contact of respondents
...
Income, religion, sex, and politics are
topics that may elicit item refusals
...
Refusals can also be specific to the method of data collection, as
in nonresponse to a mail and email questionnaires or using caller ID to screen and avoid
telephone surveys
...

The amount of effort involved in data collection is another possible way to affect
nonresponse error
...

Example: In a national telephone survey, a so-called five-day “standard” survey was compared to a
“rigorous” survey conducted over an eight-week period (Keeter, Miller, Kohut, Groves, & Presser,
2000)
...
But the two surveys produced similar results
...
Very few differences were found on substantive variables
...
Although specific respondents are individuals, organizations are not, as they
are differentiated and hierarchical
...

Tomaskovic-Devey, Leiter, and Thompson (1994) in a study of organizational response,
stated the likelihood that an organizational respondent will respond is a function of three
characteristics of the respondent:
1
...
Capacity to respond: Organizational practices and the division of labor and information affect the
assembly of relevant knowledge to reply adequately
3
...


Surrogate Information Error
In many research situations, it is necessary to obtain information that acts as a surrogate
for that which is required
...

Decision-oriented behavioral research is always concerned with the prediction of
behavior
...
Smith and Gerald S
...
Typically, researchers obtain one or more kinds of surrogate
information believed to be useful in predicting behavior
...
One may ask
about intended behavior as a means of prediction
...


Since the type of information required is identified during the problem-formulation stage
of the research process, minimizing this error requires an accurate problem definition
...

Such error can potentially arise at any stage of the measurement process, from the development
of an instrument through the analysis of the findings
...
1 depicts the stages
at which errors in eliciting information may arise when interviewing respondents for a survey
...
1 Potential Sources of Measurement Error in a Survey

In the transmittal stage, errors may be due to the faulty wording of questions or
preparation of nonverbal materials, unintentional interviewer modification of the question’s
wording, or the way in which a respondent interprets the question
...
One aspect of this regards form; form-related errors
concern psychological orientation toward responding differently to different item formats and
include:
1
...
Central tendency: reluctance to give extreme scores
3
...
217)

32
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

In the analysis stage, errors of incorrect editing and coding, descriptive summarization,
and inference can contribute substantially to measurement error
...

Experimental Error
When an experiment is conducted, the researcher attempts to measure the impact of one
or more manipulated independent variables on some dependent variable of interest, while
controlling for the influence of all other (i
...
, extraneous) variables
...
Consequently, what may be measured is not
the effect of the independent variables but the effect of the experimental situation itself
...
There are two basic approaches for handling potential
errors:
1
...
Measure or estimate the error or its impact

Minimize Error
Two different approaches can be taken to minimize total error
...
Much of
the material in Chapters 3 through 9 of this book discusses effective research methods, and as
such, involves techniques designed to minimize individual errors
...
However, this approach is often
limited by the budget allotted to a project
...
Thus, attempts to minimize one component may lead to an increase in
another
...
This means that the research designer must trade off errors when
developing a research design that minimizes total error
...
If the reduction in these nonsampling
errors exceeds the increase in sampling error, there will be a reduction in total error
...
There is a body of accepted sampling theory that allows the
researcher to estimate sampling error for a probability sample, but nothing comparable exists for
non-sampling errors
...

For individual error components, many diverse procedures can be used to estimate and
measure their impact as illustrated in Table 2
...
These are discussed where appropriate in
subsequent chapters
...
Smith and Gerald S
...
1

Selected Methods for Handling Non-Sampling Errors

Type of Error

Design to Avoid

Surrogate Information

Strive for realism

Measurement
1
...
Interviewingassociated
(e
...
, bias,
recording, cheating
in telephone and
personal
interviews)

Pretest, alternative wording,
alternative positions, etc
...


Response

4
...


Coding
(text and manually
entered data)

6
...


Analysis

Frame
Selection

Nonresponse

Select and train interviewers
correctly
Use same editor of all
interviews by one
interviewer
Use cheater questions
Use computer program to
analyze for patterns of
responses by interviewer
Randomize response
technique
Ask for verification checks
Cross-check questions
Use mail-back technique
Prepare editing manual
Train editors
Require daily return of data
Pre-code variables
Use coding manual
User computer program to
clean data
Use verification for data
entry
No remedy except
competence
Use multiple frames
Make sample element and
sample unit the same
Use probability sample
Use callbacks
Call at appropriate time
Use trained interviewers

Measure
No method of direct
measurement, as event has
not yet occurred
Experiment by using
alternative positioning, etc
...
Consequently, the error that exists for every project must be
estimated or measured
...
Although estimating or measuring errors is
better than ignoring them, there may be times when ignoring non-sampling error may not be that
bad
...
95 confidence
level only to
...
However, ignoring a non-sampling error equal in
amount to sampling error reduces the
...
83
...
Smith and Gerald S
...
Each design will have an associated expected value of information and incurred
cost
...
There are many possible ways of measuring market share of
energy drink brands, including questioning a sample of respondents, observing purchases at a
sample of retail outlets, obtaining sales figures from a sample of wholesalers, obtaining sales
figures from a sample of retailers and vending machine operators, obtaining tax data, subscribing
to a national consumer panel, subscribing to a national panel of retail stores, and, possibly,
obtaining data directly from trade association reports or a recent study by some other
investigative agency
...

The selection of the best design from the alternatives is no different in principle from
choosing among the alternatives in making any decision
...
If the design is such that
the project will yield information for solving more than one problem, the expected value should
be determined for all applicable problems and summed
...


SUMMARY
In this chapter we dealt with a subject of single most importance to the research project:
the research design
...
Finally, we
considered the errors that affect research designs
...
These chapters deal with major sources of marketing
information—respondents and experimentation—and the means of obtaining and analyzing
research information
...
Smith and Gerald S
...
F
...
The psychology experiment: An introduction to the scientific method
(2nd ed
...

Assael, H
...
(1982, Spring)
...
sampling errors in survey research
...

Feldman, J
...
Considerations in the use of causal-correlational technique in applied
psychology
...

Fern, E
...
(1982)
...
Journal of Marketing
Research, 19 (February), 1–13
...
(2003)
...
Thousand Oaks, CA: Sage
...
, Miller, C
...
, Groves, R
...
, & Presser, S
...
Consequences of reducing
non-response in a national telephone survey
...

Lilien, G
...
, & Searls, K
...
How errors add up
...

Monroe, K
...
, & Petroshius, S
...
(n
...
Developing causal priorities
...

Tomaskovic-Devey, D
...
, & Thompson, S
...
Organizational survey response
...

Tull, D
...
, & Albaum, G
...
(1973)
...
New York: Intext
Educational Publishers
...
, Albaum, G
...
(2003)
...


36
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 3
SECONDARY SOURCES OF INFORMATION
“A few months of field work collecting data can save several hours in the library”

Secondary data (or secondary information) is information that has been collected by
persons or agencies for purposes other than the solution of the marketing research problem at
hand
...
The key point is that the data were collected for some other project, or
reason, than the current one
...

Primary data is often collected from a respondent, an individual who provides information either
passively through the observation of his or her behavior, or actively through verbal response
...

In addition to primary and secondary data, there exists commercial data sold in the form
of syndicated services
...
Since
these data relate to ongoing concerns of a marketer they can be viewed as primary data
...
Thus, there are elements of secondary data
...

In this chapter, we discuss the reasons for obtaining secondary information, types of
secondary information, sources of external secondary data, and syndicated services that provide
commercial data
...
Secondary research can help
provide a clearer picture of a problem so that researchers and managers can make the necessary
critical decisions
...
This search should be conducted early in the problem investigation stage
and prior to any organized collection of information from primary sources
...

Secondary Information May Solve the Problem
If adequate data are available from secondary sources, primary data collection will not be
required
...
based a long-running advertising campaign on the
theme “soup is good food
...
Campbells
(1959: youtube
...
com/watch?v=4PtZYlcyXfI )

Secondary Information Search Costs Substantially Less
A comprehensive search of secondary sources can almost always be made in a fraction of
the time and cost required for the collection of primary information
...
Searching for secondary
37
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

research helps you avoid duplicating primary research and optimizes research expenditures by
acquiring only information that cannot be found elsewhere
...


Secondary Information Has Important Supplementary Uses
When secondary information cannot solve the research problem, it can provide invaluable
supplemental uses:
1
...
The analysis of available
secondary data will almost always provide a better understanding of the problem and its context,
and will frequently suggest solutions not considered previously
...


Planning the collection of primary data
...
It
may also be of value in establishing classifications that are compatible with past studies so that
trends may be more readily analyzed
...


Defining the population and selecting the sample
...


The researcher must be careful when using only secondary data
...
Often, little is known about the
reliability of secondary research studies
...

Example: One company wants to do a segmentation analysis on foreign markets with a particular
emphasis on examining demographics (Albaum, Duerr, & Strandskov, 2005, Chap
...
The
company is considering using the official government census of the population
...

What one gets from a census depends on what was on the census form in the first place—
typically a mixture of traditional questions and new items of interest to public policy makers and
civil servants at the time
...
S
...
Income is one of the
major dimensions of U
...
segmentation research, but many highly developed nations ignore the
income question in their censuses
...
Moreover, the data may not use the same categories when showing relevant
distributions of demographic variables, such as age
...


38
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

TYPES OF SECONDARY INFORMATION
Secondary information falls into two categories, data that is available within the company
(internal data) and that which must be obtained from outside sources (external data)
...
Orders
are received and filled, costs are recorded, warranty cards are returned, salespeople’s reports are
submitted, engineering reports are made—all are collected for other purposes, but may be useful
to the researcher (Andreasen, 1988, pp
...
The key, of course is knowing where they are
and how to access them
...

Example: Spectra Physics Lasers Division (producing laser grocery store scanners) regularly
performs customer satisfaction studies
...
Also, they can be secondary data to Retail Systems should they be used at a
much later date for aiding in decision-making or for purposes other than those originally
intended when the studies were done
...
It
also is applicable to all of the major types of marketing research projects and is mainly
concerned with the noncontrollable aspects of the problem:
Total market size
Market characteristics
Competitor products, prices, promotional efforts, and distribution methods

As an example, a consumer goods company is considering whether it should establish a
direct selling operation
...
The Direct Selling
Association (DSA) provides secondary information in the form of a regular survey of the
industry
...
Some of the types of industry statistics are the following:
Estimated U
...
sales
Estimated U
...
salespeople
Percent of sales by major product groups
Location of sales
Percent of sales by census region
Sales strategy
Compensation structure by percent of firms
Compensation structure by percent of sales dollars
Compensation structure by percent of salespeople

39
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The following are some of the sales force demographics available:
Gender
Age
Education
Independent contractor/employee status
Hours per week dedicated to direct selling
Average time spent on direct selling tasks
Main reasons for becoming a direct sales representative
Percent of salespeople by distributorship type

Age, education, average time spent on direct selling tasks, and main reason for becoming
a direct sales representative data are from DSA’s National Sales force Survey
...
dsa
...


SOURCES OF EXTERNAL SECONDARY DATA
The major original sources of external secondary information are:
1
...

3
...

5
...
g
...
Both
governmental and trade sources are so important that the experienced researcher will be
thoroughly familiar with them in his or her field of specialization
...

Commercial services of many types are available that are highly useful for specific research
problems
...
Such data are usually available on a reasonably
recent basis from censuses conducted by federal, state, local, and, when needed, supranational
governments
...
census
...
This reference abstracts
data from original reports and gives some useful material on social, political, and economic
matters
...

The State and Metropolitan Area Data Book is a publication of the Bureau of the Census
that is available online in PDF format
...
It covers information about
numerous topics relevant for both B2C and B2B marketing, including population, income, labor
force, commercial office space, banking, health care, housing, and so forth
...
census
...
html
40
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The Census of Population http://www
...
gov/population/www/ and the Census of
Housing http://www
...
gov/hhes/www/housing
...
S
...
Updates of various
census measurements based on smaller yearly surveys are available in Current Population
Reports http://www
...
gov/main/www/cprs
...
census
...
html
...

Data from the U
...
Census Bureau is available online for custom data analysis, on CDROM, and in report form as downloadable PDF files
...
The company previously mentioned, GeoLytics (http://geolytics
...

Demographic reports (and maps)
– Custom data sets and reports
– Area segmentation
– Area-to-area correspondence files
– Banking and realtor tract level data and maps
Services
– Geocoding (GPS) addresses
– Custom-built databases
– Normalized data for across census comparisons

Other companies include census data in mapping software that is used for geographic
market analysis
...
One supplier is
Scan/US, Inc
...
scanus
...
S
...
includes maps for the entire United States that include all types of demographics
...

To illustrate, SRDS ( www
...
com/frontMatter/ips/lifestyle/index
...
This annual provides demographic and lifestyle information for 210 Designated
Market Areas (DMAs) in the United States
...
1 and Figure 3
...

Most popular lifestyles for each DMA: Specify a lifestyle and then identify what other interests
frequently appeal to those consumers and what demographic information corresponds to that
profile
...
Smith and Gerald S
...
1

Demographics and Lifestyles, Seattle-Tacoma, Washington

42
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

43
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 3
...
S
...
srds
...
jpg
Nielsen’s Claritas division, a provider of solutions for geographic, lifestyle and behavioral target
marketing, has developed a demographic widget that is available as a free download for personal
electronics http://www
...
com/target-marketing/nielsen-claritas-demographic-widget
...


Market size studies (e
...
, size in sales dollars or units) often are conducted by trade
associations, media, firms in the industry, and private research organizations
...
Industry-type studies may be concerned with
such types of information as total market size, market characteristics, market segments and their
size and characteristics, and similar types of information
...
conducts a single-source continuing survey, primarily
aimed at the advertising industry that provides demographics, lifestyles, product usage, and
exposure to all advertising media data
...

http://www
...
com

Information on new products and processes is available from such sources as patent
disclosures, trade journals, competitors’ catalogs, testing agencies, and the reports of
governmental agencies, such as the Food and Drug Administration, the Department of
Agriculture, and the National Bureau of Standards
...

Publishers Information Bureau provides a compilation of expenditures by medium for each
competitor
...
magazine
...
aspx
Audit Bureau of Circulations provides data on the numbers of magazine copies sold under
specified conditions
...
accessabc
...

http://www
...
com/portal/main?action=LinkHit&frameset=yes&link=ips

Mediamark Research, Inc
...

http://www
...
com

44
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Arbitron Radio and Television Reports (http://www
...
com/home/content
...
acnielsen
...
shtml,
Starch Advertising Readership Service http://www
...
com/services
...

EXHIBIT 3
...

Each year, Starch measures over 25,000 ads in over 400 magazine issues
...
Then the data are put
into a context: The ad is ranked not only against other ads in the issue but also against other ads in its
product category over the past two years
...


The Benefits of Starch Ad Readership
In-Depth Analysis
Campaign analyses inform clients not only about the scores of the ads but also why they
performed as they did and what can be done to improve the ads
...
g
...


Extra Questions
To give you information on advertising likeability, persuasiveness and intent to purchase
Many times, if you ask a publisher to Starch an issue your ads will appear in, they will assume
the cost and pass on the data to you for free
...


45
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Internet Databases
The Internet has become the staple of research and provides access to most commercial
electronic databases
...
dialog
...
lexisnexis
...
dowjones
...
htm
...
Bibliographic databases that index publications
2
...
Statistical databases of demographic, econometric, and other numeric data for forecasting and
doing projections
4
...
Full-text databases from which an entire document can be printed out
...
All that is needed is personal
computer with internet access or a CD-ROM/DVD
...
Database
marketing has been defined as an extension and refinement of traditional direct marketing, but
uses databases to target direct response advertising efforts and tracks response and/or
transactions
...
Data may come
from secondary and/or primary sources
...
APIs can be used to link and integrate data from multiple sources in real
time
...
Databases can be used to estimate market size, find segments and
niches for specialized offerings, and even view current customer use and spending (Morgan,
2001)
...

Today, data mining is in high demand as a research focus
...
Viewed as a complement to more traditional statistical techniques of
analysis, two of the more powerful data mining techniques are neural networks and decision
trees (Garver, 2002)
...


46
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

EXHIBIT 3
...
The plan will outline all the key areas of inquiry and will provide
a systematic pathway to search, retrieve, and arrive at the desired data, information, and
knowledge
...
One of the best
ways is to do a knowledge inventory or review to determine what you already know or have in
your possession, both primary and secondary research
...
Start by applying the learning from your knowledge inventory and specify the
new areas that are critical to your project
...

Search across multiple sources
...
You only rarely will find what you need in one place
...
Use techniques of trend analysis to identify
patterns that emerge when various information elements are combined; for example, content
analysis, stakeholder analysis, paradigm shift, trendlines, critical path analysis, sector analysis
(technological, social/cultural, occupational, political, economic), or other analytic techniques
that facilitate integration of diverse data and information and identification of underlying
patterns
...
Many of the professional online database producers and vendors offer thousands
of full-text articles and resources that contain analyses
...

Enhance the robustness of your data or information through multiple-source validation
...


SYNDICATED SERVICES
Some of the aforementioned commercial services are examples of what are called
syndicated services
...
In general, syndicated data are made available to all who wish to subscribe
...
Since
these data are not collected for a particular firm for use in a specific research problem situation,
they can properly be viewed as secondary data
...

The syndicated Survey of American Consumers (based on surveying more than 25,000
adults) by Mediamark Research, Inc
...
This survey
provides data useful for detecting a marketer’s best prospects by providing answers to such
questions as the following:

47
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

How many customers are there for the products or services we market? Is the size of the market growing?
Stabilizing? Or shrinking?
Who are the customers? How old are they? What do they earn? Where do they live?
How do customers differ in terms of how often and how much they buy? Who are the heaviest purchasers
of the product?
What brands are customers buying? How have shares of the market changed? What differences are there
among brand buyers?
What’s the best way of reaching prospects? Which media vehicles and formats are most efficient in
delivering the message to the customer?

Mediamark is able to profile American consumers on the basis of more than 60
demographic characteristics and covers usage of some 500 product categories and services and
6,000 brands
...

One of the most widely used approaches is the continuous panel, which refers to a sample of
individuals, households, or firms from whom information is obtained at successive time periods
...
As consumer purchase panels, which record purchases in a consumer diary and submit them
periodically
...
As advertising audience panels, which record programs viewed, programs listened to, and
publications read
...
As dealer panels, which are used to provide information on levels of inventory, sales, and prices
...

These types of panels furnish information on at regular intervals on continuing purchases of the
products covered
...

Data are recorded in diaries, either online or are mailed in each month
...
npdor
...
This panel comprises 13,000 families and is national in coverage
...
Other well-known national consumer
panels are maintained by Synovate (synovate
...
com) ACNielsen
(http://en-us
...
com), and IRI (infores
...

Advertising audience panels are undoubtedly more widely publicized than other panels
...
These panels are
operated by independent research agencies rather than the media—both for reasons of economy
and to avoid any question of partisanship
...
Smith and Gerald S
...

The type of activity is recorded automatically; household members merely have to indicate their
presence by pressing a button
...
In local markets, the
sample may be 300 to 400 households
...
Mediamark’s national survey and IRI’s BehaviorScan are examples of such singlesource data
...

The information obtained from the types of syndicated services described previously has
many applications
...
Trends and
shifts in market composition may be analyzed both by type of consumer and by geographic
areas
...

Analyses of trends of sales by package or container types may be made
...
Trends in competitor pricing and special promotions
and their effects can be analyzed along with the effects of the manufacturer’s own price and
promotional changes
...
Similarly, innovative buyers may be identified for new products and an analysis of
their characteristics made to aid in the prediction of the growth of sales
...
One reported use of this syndicated
service has been to design products for specific segments
...


SUMMARY
This chapter has been concerned with secondary information and sources of such
information
...
Then, various sources and types of secondary information—internal
and external—were discussed in some depth
...


49
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

REFERENCES
ACNielsen: http://en-us
...
com
ACNielsen – Claritas Demographic Widget, http://www
...
com/target-marketing/nielsenclaritas-demographic-widget
...
, Duerr, E
...
(2005)
...
Harlow, UK: Pearson Education
...
(1988)
...
Homewood, IL: Dow-Jones Irwin
...
, & Linoff, G
...
Data mining techniques: For marketing, sales, and customer
support
...

Berry, M
...
(2002)
...
New York:
Wiley
...
youtube
...
youtube
...
, & Hancock, M
...
Data mining explained: A manager’s guide to customer
centric business intelligence
...
dialog
...
dsa
...
dowjones
...
htm
Elkind, F
...
(1995, July 31)
...
Marketing News, 29, 2ff
...
S
...
Try new data mining techniques
...

IRI: infores
...
lexisnexis
...
S
...
Research boosts database’s power
...

NFO: mysurveys
...
npd
...
Smith and Gerald S
...
http://www
...
com
RoperASW
...
Available at RoperASW
...
Accessed January 9, 2002
...
srds
...
html
SRDS Graphical Reports,
http://www
...
com/frontMatter/ips/lifestyle/images/LMAA_overview-with-map
...
com
U
...
Census Bureau, Statistical Abstract, http://www
...
gov/compendia/statab/
U
...
Census of Population, http://www
...
gov/population/www/
U
...
Census of Housing, http://www
...
gov/hhes/www/housing
...
S
...
census
...
html
U
...
Current Construction and Housing Reports, http://www
...
gov/prod/www/abs/conshou
...
Smith and Gerald S
...
1
Selected Secondary Sources
Marketing/Advertising
Directories
American Marketing Association
...
Annual
...
greenbook
...
Foreign as well as U
...
research
firms are entered alphabetically in a single list
...
Annual
...
bowker
...
php/component/content/article/2/33
In addition to its radio section, this source includes sections that survey the broadcasting industry, profile television
and cable stations, and focus on programming, professional services, technology, advertising and marketing, and
satellites
...
Annual
...
dirmktgplace
...
); service firms and suppliers to direct
marketing companies (printers, list brokers and managers, computer services, etc
...
); and lists of organizations, courses, awards, and so on
...

Rand McNally
...
Annual
...
randmcnally
...
In addition to a series of area maps
highlighting major military installations, trading areas, retail sales, and manufacturing centers, it includes statistics
on business and manufacturers, retail trade, sales, and the largest corporations in the United States
...
Annual
...
redbooks
...
asp
Also known as the Agency Red Book, this directory contains current information about 4,500 American advertising
agencies and their branches, including their names and addresses, specialization, major accounts, and key staff
members
...
Annual
...
It focuses on companies and other organizations
that advertise and includes trade names, the types of media used for advertising, the advertising agencies employed,
and, frequently, annual advertising budgets
...
http://www
...
com/en-US/catalogs/databases/detail/abi_inform
...
Over 800 journals are indexed and each article is fully
annotated
...
Current five years on disc
Business Index
...

This is a microfilm index that is a cumulative, three-year index to over 800 business periodicals and newspapers
...
Monthly
...
ovid
...
jsp
Indexes over 300 business periodicals
...

Gale Group F & S Index United States
...

http://www
...
cengage
...
Indexes over 750 periodicals, reports, and trade journals
...
Smith and Gerald S
...
Semimonthly
...
Over 1,400 journals are indexed
...

Consumer Index
...

Topicator
...

Other Sources
New York Times Index
...
nytimes
...
html
Statistical Indexes
1
...
Monthly
...
lexisnexis
...
aspx?pid=65&type=&parentid=58
The ASI is a master guide to statistical publications of the federal government
...
It is arranged by subject with references to microfiche reproduction of the
publication found there
...
Statistical Reference Index
...

Same format as aforementioned ASI except it indexes and abstracts statistics contained in publications not issued by
the federal government
...

Government Publications
1
...

Records and indexes documents received by the Government Printing Office from all areas of government,
legislative, judicial, and executive branches, and independent and regulatory agencies as well
...
Congressional Information Service
...
Monthly
...

Bibliographies
1
...
What’s New in Advertising and Marketing
...

List by subject new books, periodicals, and trade publications in marketing and advertising
...

Special Marketing, Advertising, and Consumer Reference Sources
Editor and Publisher Market Guide
...
http://www
...
com/eandp/resources/market_guide
...
S
...

Leading National Advertiser
...
$ Summary
...

Lists brands alphabetically and shows total 9-media expenditures, media used, and parent company
...

Sales and Marketing Management
...

1
...
Annual http://www
...
com/sbponline/index
...
Also lists areas by population, number of households, total retail sales, and much more
...
Survey of Buying Power Data Service
...

Has more detailed and additional information than the Survey of Buying Power
...
Smith and Gerald S
...
http://www
...
uga
...
html
Forty volumes devoted to the determination of use of various products and services
...
http://www
...
com/portal/main?action=LinkHit&frameset=yes&link=ips
This service offers separate directories giving advertising rates, specifications, and circulation for publications,
broadcast stations, and so forth in the following media: business publications (monthly), community publications
(semiannual), consumer magazines and agri-media (monthly), newspapers (monthly), print media production
(quarterly), spot radio (monthly), spot television (monthly), and weekly newspapers (monthly)
...
Entries include daily and Sunday circulation figures, by county, for each newspaper
...
Million Dollar Directory
...

Lists over 160,000 American companies having an indicated net worth of $500,000 or more
...
Annual
...
S
...

Directory of Corporate Affiliations
...

Each directory features lists of divisions and subsidiaries of parent companies
...
Various call numbers
...

Includes Bank and Finance, Industrial, Municipal and Government, Over-the-Counter, Public Utility,
Transportation, and International
...

Value Line Investment Survey
...

Walker’s Manual of Western Corporations and Services
...

Standard and Poor’s Corporation
...

This loose-leaf service includes for 22 industry categories detailed analysis of each category and of the industries
that comprise it
...
Quarterly
...

U
...
Industrial Outlook
...

Contains short-term forecasts for specific industries
...

Indexes and Guides
Predicasts Funk and Scott Indexes
...
F & S Europe
...

Covers articles or data in articles on foreign companies, products and industries
...
F & S International Index
...

Covers articles or data in articles on foreign companies, products, and industries
...
Smith and Gerald S
...
Annual
...

Europa Yearbook
...

Includes history, economic affairs, economic statistics, constitution, government, and so on of individual countries
...
Semiannual
...
S
...

Organization for Economic Cooperation and Development Economic Surveys of the O
...
C
...

Annual
...

Overseas Business Reports
...

Provides current and detailed marketing information for businesses evaluating the export market
...

Price Waterhouse Guide Series
...

A series of guides on various aspects of doing business where Price Waterhouse has offices or has business contacts
...
Over
75 countries are represented in this series
...
Annual
...
Analysis of inflation, rates of interest, exchange rates, trade balances, commodity prices, and indebtedness
are included
...

Directory of American Firms Operating in Foreign Countries
...
S
...

Directory of Foreign Firms Operating in the U
...

Has:
1
...

2
...

3
...

Dun and Bradstreet
...
Annual
...

Foreign Commerce Handbook
...

Information on all phases of international business is included in this reference source, including foreign-trade
services, daily language of foreign commerce, bibliography, and lists of organizations
...
Annual
...
Smith and Gerald S
...
S
...
S
...

Worldwide Chamber of Commerce Directory
...

Listing of U
...
Chambers of Commerce, American Chambers of Commerce abroad and chief executive officer, and
Foreign Chambers of Commerce in principal cities of the world
...
Monthly
...
Weekly
...

1
...
Weekly
...
These newsletters cover the following areas: Asia, China, Eastern Europe, Europe,
International, and Latin America
...
Financing Foreign Operations
...

Current guide to sources of capital and credit in 34 major markets
...
Investing, Licensing and Trading Conditions Abroad
...

Covers African-Middle East, Europe, Asia, North America, and Latin America
...

4
...
Irregular
...
I
...
“Marketing in China,”
“Andean Common Market,” and so on
...
Worldwide Economic Indicators
...
Includes G
...
P
...

Commerce Clearing House
...
Weekly
...
International Business Series
...

Very much like Price Waterhouse Guides described previously but not as extensive in coverage
...
Really Useful (Web) Sites for International Trade Professionals
...
org, newsletter@fita
...
Twice weekly
...

Statistics
Demographic Yearbook
...

Comprehensive collection of international demographic statistics
...

U
...
Statistical Office
...
Annual
...
Comprehensive compendium of international comparable data
for the analysis of socioeconomic development at the world, regional, and national levels
...
Yearbook of Labor Statistics
...
Smith and Gerald S
...

U
...
Department of Economic and Social Affairs
...
Annual
...

U
...
Statistical Office
...
Annual
...
Statistical Yearbook
...

Contains tables grouped according to various subjects: population, education, libraries and museums, book
production, newspapers and broadcasting, television, and cultural expenditure
...

U
...
Statistical Office
...
Annual
...
The United Nations has
separate Economic and Social Commissions for each geographical region
...
The titles of these annual reports are listed next, followed by the title of the periodical and
supplements that update them if one exists
...
International Financial Statistics
...

Data on exchange rates, international reserves, money and banking, trade, prices, and production for all IMF
member countries
...
International Monetary Fund
...
Annual
...
Includes statistics for goods, services,
capital, SDRs, and so on
...

This appendix was developed from information provided by Rodney Christensen, former
reference librarian, Knight Library, University of Oregon
...
Smith and Gerald S
...
Smith and Gerald S
...
We can
define an interview as a form of dyadic (person-to-person) communication that involves the
asking and answering of questions
...
g
...
), psychological attributes (e
...
, perceptions,
attitudes, expectations, motives), and behavior (e
...
, errors in asking questions, probing,
motivating, recording responses)
...
These factors have a
direct influence when an interactive interview is conducted and are implied in self-report
interviews (e
...
, mail, e-mail)
...
The respondent’s role
expectations, the interviewer’s task behavior, differences in social desirability of response
alternatives, extent of topic threat, and salience of topic all affect the responses received
...
For example, styles can range from socio-emotional
(maintaining a warm, sympathetic, and understanding relationship with the respondent) to formal
(where the person-oriented actions of the interviewer are held to a socially acceptable minimum)
...

EXHIBIT 4
...

Involvement encompasses the degree to which each party wants to take part in the
interview, including the degree of commitment of each to making it a success
...

Relationship is the degree of warmth or friendship between the interview parties
...
Context
...
This also includes status
differences between parties, temperature, privacy, and time
...
Content
...
It involves topic selection
and treatment, arguments, supporting materials, language, and questions and answers
...
Structure
...

59
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

4
...
The willingness on the part of both parties to reveal their “true” selves to
one another
...
Feedback
...
g
...
Cooperation
...

7
...
The potential or actual struggle between parties because of incompatible or
opposing needs, desires, demands, and perceptions
...
Trust
...

Involvement, Control and Relationships have some effect upon each of the elements
...
Although they are independent of each other, they have strong
interdependence as well
...
J
...
B
...
© 1985
William C
...
Reprinted with permission of The McGraw-Hill Companies, pp
...
An interview is either
structured or unstructured, depending on whether a formal questionnaire has been formulated
and the questions asked in a prearranged order
...
Crossclassifying these two characteristics helps us to identify four different types of interviews
Objective Interviews (a) structured and direct
(b) unstructured and direct
Subjective Interviews (c) structured and indirect
(d) unstructured and indirect
...
We discuss each type
of interview in turn (although the discussion of the two indirect types of interviews is combined)
...

Structured-Direct Interviews
Structured-direct interviews are the usual type of consumer survey to “Get the Facts” and
obtain descriptive information
...

Example: A marketing research manager of a bedroom-furniture manufacturer wants to find out
how many and what kinds of people prefer various styles of headboards and dressers
...
The resulting interview is structureddirect in nature
...
Smith and Gerald S
...
Since the questions are
formulated in advance, all the required information can be obtained in an orderly and systematic
fashion
...
Pretests can (and should) be made
on the questionnaire to discover any problems in the wording or ordering of questions before the
questionnaire is finalized
...
The interviewer’s role is simply to ask questions
...
Smith and Gerald S
...
This provides maximum control of the interviewing process
and reduces the variability in results caused by differences in interviewer characteristics
...
The
standardized, direct questions allow for uniform recording of answers, thereby reducing errors in
editing, tabulating, and analysis of the information
...
The structured-direct interview is by far the most
commonly used type of interview in marketing research
...
2
...
2 Structuring Conversational Interviewing
Despite pretesting, every survey question contains terms that have the potential to be
understood differently than the survey designer intends
...
What exactly do you
mean by anti-inflammatory?” The interviewer is now faced with a choice
...
Interviewers must read exactly the same question
and never interpret the question in any way (Fowler, 1991; Fowler & Mangione, 1990)
...

Another school of thought holds that the interviewer in the example above should help the
respondent and define “anti-inflammatory
...
An approach suggested is that interviewers be
allowed to use conversationally flexible interviewing techniques
...

Online surveys have the ability to provide standard definitions through context sensitive help
...
To the degree that respondents needing help would
receive it, response validity would increase, but the interaction with the respondent would be
standardized
...
In the unstructured-direct method of interviewing, the
interviewer is given only general instructions to obtain the type of information desired
...

Many research projects go through an exploratory phase in which researchers contact
respondents and hold unstructured interviews
...
This type of interview
is also often useful for obtaining information on motives
...

62
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

To use the bedroom furniture example again, if the owner of a bedroom set is asked the
free-answer question, “Why did you buy your bedroom set?” the answer is almost certain
to be incomplete, provide proximate causes and may be worthless
...
” When
motivations are given, such as “We enjoy a comfortable mattress that gives us a good
night’s sleep,” they are rarely complete
...
In addition, it is probable that motives other
than “enjoyment” influenced the purchase
...
The interviewer will continue to ask probing questions: “What did you mean by that
statement?” “Why do you feel this way?” “What other reasons do you have?” The interviewer
continues with similar questions until satisfied that all the information that can be obtained has
been obtained, considering time limitations, problem requirements, and the willingness and
ability of the respondents to verbalize motives
...
The interview may be conducted in a seemingly casual, informal manner in which the
flow of the conversation determines which questions are asked and the order in which they are
raised
...
The flexibility inherent
in this type of interview, when coupled with the greater informality that results when it is
skillfully used, often results in the disclosure of information that would not be obtained in a
structured-direct interview
...
The
unstructured interview can therefore be only as effective in obtaining complete, objective, and
unbiased information as the interviewer is skilled in formulating and asking questions
...
Higher per-interview costs result, both as a result of this requirement and
the fact that unstructured interviews generally are longer than those that use a questionnaire
...

Structured-Indirect and Unstructured-Indirect Interviews
A number of techniques have been devised to obtain information from respondents by
indirect means
...
Many of these
techniques employ the principle of projection, in which a respondent is given a non-personal,
ambiguous situation and asked to describe it
...
The description,
therefore, involves a projection of personality characteristics to the situation described
...


63
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

REDUCING RESPONSE AND NONRESPONSE BIAS
A major concern of the research planner when choosing which interview medium to use
is the potential systematic error (i
...
, bias) that might arise
...
At this point, it is useful to explore how to reduce the nonsampling based error that occurs during the interview process
...
The following sections will discuss these topics
...
There are two types of inaccuracies: predictive and concurrent
...
Or, alternatively, he does not now intend to buy, he answers “No” to
the question, and then buys a car within the six-month period
...
This situation is a predictive
inaccuracy
...

Concurrent inaccuracy occurs when the respondent intentionally does not provide
accurate information
...

For our car purchase example, suppose the respondent answers “Yes”, but really has no
intention of buying a sports car within this period or, conversely, answers “No”, but does
intend to buy a sports car
...
The intention is to provide inaccurate information
...
Concurrent inaccuracies may also apply to instances where
observation is used; the observer charged with reporting the events or behavior may be unable or
unwilling to provide the desired information
...


64
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Inability to Respond
Even such a simple and straightforward question as “What is the model year of your
family car?” may result in an information-formulation problem, particularly if the car is several
years old
...
Finally, if respondents were asked, “What reasons did you have for buying Brand “A”
tires instead of some other brand?” most respondents would have even more difficulty in
providing an accurate answer
...
Often time since an event occurred
(purchase) are underestimated or overestimated
...
For instance, in a personal or telephone interview survey, follow-up calls are often
not made to confirm the answers given
...
A question (or even an entire questionnaire) may be unrealistic, deficient,
or directed to the wrong persons
...
Poorly defined terms
or words with different meanings can lead to inaccurate, or even deliberately falsified responses
...

In addition, the following items also create inaccuracies through the inability to respond
accurately:
Telescoping
Questions that ask respondents to reconstruct past experiences run a high risk of response
bias
...
In a study of durable-goods purchases in the
United States, respondents on average displayed forward-telescoping biases (reporting something
happened more recently than it did), and the magnitude of this reporting increased (Morwitz,
1997)
...

Telescoping can be reduced by using bounded recall procedures, which involve asking
questions about the events of concern in previous time periods as well as the time period of
research interest (Sudman, Finn, & Lannam, 1984)
...


65
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Exhibit 4
...

Important/Urgent questions are a first priority and easy to identify
...
It is the other two categories that can cause most of the
problems
...
In the short run, a company will carry on without answers to Important/Non-urgent
questions
...


Unwillingness to Respond
When we move to the problem of unwillingness of respondents to provide accurate
information, the topic is more complex
...

Except in those instances where the respondent provides information by being observed
in a natural situation, there are always costs (negative utilities) attached to his or her formulating
and sharing information
...

Investigator Expectations
A complex source of inaccuracy in response stems from the respondents’ appraisal of the
investigator and the opinions and expectations imputed to him or her
...
In this case, one
exceptionally well-dressed, carefully groomed, competent interviewer conducted all of
the interviews
...
The reported
brands of cosmetics used were much less expensive, in this series of interviews
...
This happens when an interviewer finds a particular question too embarrassing to
ask; when the interview finds it easier to self complete the survey forms rather than conduct the
interviews; or when interviewers have friends complete the survey
...

Interviewer cheating can be kept to a low level of incidence but not eliminated
completely
...
Smith and Gerald S
...
In addition, control procedures can and should be established to reduce it even
more
...
If the
information on an initial interview is found to disagree significantly with that on the call-back
interview, additional call-backs may be made on respondents originally interviewed by the same
person
...

Other control procedures include the use of “cheater” questions and the analysis of
response patterns
...
Likewise, the analysis of patterns of
responses for interviewer differences will disclose interviewer cheating when significant
variations from expected norms occur
...

Time Costs
Perhaps the most common reason for respondent unwillingness to provide accurate
information, or any information for that matter, is the result of the time required to make the
information available
...
When possible to do so, a respondent will tend to
act in a manner that will reduce time costs
...

When conducting telephone and personal interviews the interviewer might ask “Is this a
good time to answer some questions, or would you rather set a time when I could contact you
again?” Experience has shown this latter technique only slightly lowers response rates
...
All researchers experience difficulty both in
recognizing the items that demand prestige content, and in measuring the resulting amount of
inaccuracy
...

An example of a still more subtle prestige association occurred in a study on nationally known
brands of beer
...
Since sales data indicated a strong
preference for regular beer, it was evident that the information was inaccurate
...
They had, therefore, given answers that, in their view, were associated
with a higher level of prestige
...
One solution to this problem is to
ask for the information in two different ways
...
In a follow-up study, when they were asked, “Do you think your
neighbor is afraid to fly?” (a technique known as the third-person technique), most of the
neighbors turned out to have severe anxieties about flying
...
Smith and Gerald S
...

When such is the case, both nonresponse and inaccuracy in the responses obtained can be
anticipated
...
It
should be recognized however, that invasion of privacy is an individual matter
...

The investigator should attempt to determine sensitivity if it is suspected to be a problem
...
A comprehensive treatment of sensitive information
and how to ask questions about it is given by Bradburn and Sudman (1979)
...

Ambiguity, therefore, occurs in the transmission of information, through either communication
or observation
...
Unambiguous communication in research requires
that the question asked and the answers given each mean the same thing to the questioner and the
respondent
...
If the question is not clearly
understood by the respondent, frequently the answer will not be clearly understood by the
questioner
...


Even a careful reading of these two questions may not disclose any real difference in
their meaning
...
In
pretesting, however, it was discovered that some housewives answered “Yes” with other referent
in mind
...
If the wording of the question had not been changed, there would have been a
complete misunderstanding in some cases of the simple, one-word answer “Yes
...
Examples are, “How
come?”, “What?”, and “How?” This “elliptical sentence” requires the respondent to first consider
the context of the sentence and then add the missing parts
...

68
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The understanding of questions is an issue that goes beyond ambiguity
...

Most personal and telephone interviewing uses standardized interviewing, meaning that the
interpretation of questions is left up to the respondent
...
2, one
interesting approach taken in online surveys by Qualtrics
...

Procedures for Recognizing and Reducing Ambiguity in Communication
Every research design that uses communication to obtain information should have as
many safeguards against ambiguity as possible
...

Three procedural steps are useful for these purposes and should be considered in every
project:
1
...
Pretesting
3
...
Alternative question wording
...
In questions where ambiguity is
suspected, it is advisable to consider alternative wordings and forms of questions to be asked of
sub-samples of respondents
...
In personal and telephone
interviewing situations the interviewers can likewise be instructed to change the order for onehalf the interviews
...

2
...
Pretesting of questionnaires is a virtual necessity (Converse & Presser,
1986, pp
...
The only way to gain real assurance that questions are unambiguous is to try
them
...
To be truly
effective, however, pretesting of questions should be conducted by asking them of a group of
respondents who are similar to those to be interviewed in the final sample
...
Be sure to
ask about the exact nature of the problem
...
It is the rule, rather than the exception, that
questions will be revised as a result of pretesting
...

3
...
Information obtained through communication should be
verified by observation whenever cost, time, and the type of information desired permit
...
For example, a housewife may
object to a pantry audit to verify that brands she indicated as preferred are on hand
...
Smith and Gerald S
...
In making observations we each select, organize, and interpret
visual stimuli into a picture that is as meaningful and as coherent to us as we can make it
...

As an illustration, a cereal manufacturer ran a promotional campaign involving a drawing
contest for children
...
The contest was run, the
prizes awarded on the basis of artistic merit, and the brand manager turned his attention to other
matters
...
He found that a sizable proportion of them showed a child eating cereal
alone, often with no other dishes on the table
...
Later studies by the company’s marketing research
department showed that cereals are eaten between meals by children in greater amounts than are
eaten for breakfast
...

Nonresponse Error
A nonresponse error occurs when an individual is included in the sample but, for any of
many possible reasons, is not reached or does not complete the survey
...

Non-response errors differ in nature depending on the mode of survey administration
...
They may be
away from home during the day and differ from those in which at least one member can usually
be found at home with respect to age, number of small children, and the proportion of time in
which the wife is employed
...

Internet surveys have the potential of increasing contact because the survey invitation
appears in the inbox awaiting the potential respondent’s reply
...

The seriousness of nonresponse error is magnified by the fact that the direction of the
error is often unknown
...
These additional respondents were
shown to provide more centrally distributed responses, thus producing no additional error
...
Smith and Gerald S
...

Evangelista, Albaum and Poon, (1999) suggest that four general motivations drive survey
response
...
4 suggests that response rates may be increased (and non-response bias
decreased) by using specific motivational techniques and inducements
...
4 Theories of Survey Response
Why do people participate as respondents in a survey? The question is often asked by marketing
researchers, perhaps all too often implicitly, and seldom is an answer provided other than in terms of
specific techniques (including inducements) that have been used to increase participation
...

Exchange
The process of using survey techniques to obtain information from potential respondents can be viewed
as a special case of social exchange
...
Whether a given behavior occurs is a function of the perceived costs of engaging in that
activity and the rewards (not necessarily monetary) one expects the other participant to provide at a later
date
...
The costs for responding must be minimized
...
The rewards must be maximized
...
There must be a belief that such rewards will, in fact, be provided
...
As used to
explain survey response, the theory postulates that reducing dissonance is an important component of
the “respond/not respond” decision by potential survey respondents
...
Assuming
that failure to respond might be inconsistent with a person’s self-perception of being a helpful person, or
perhaps at least one who honors reasonable requests, failure to respond will produce a state of
dissonance that the potential respondent seeks to reduce by becoming a survey respondent
...
Delaying a decision, therefore, may in itself be a dissonance-reducing response
...
Interpretations are made on the basis of selfobservation
...
These attitudes (selfperception) then affect subsequent behavior
...
To increase the
precision of this paradigm, the concepts of salience (behaviors one has attended to), favorability (the
affect or feeling generated by a given behavioral experience), and availability (information in memory) are
utilized
...
Labeling involves classifying
people on the basis of their behavior such that they will later act in a manner consistent with the

71
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

characterization
...

Commitment and Involvement
Of concern here is the range of allegiance an individual may be said to have for any system of which he or
she is a member
...
Persists over some period of time
2
...
Rejects other acts of behavior
Consequently, the major elements of commitment are viewed as including the following:
1
...

2
...

3
...

A person who is highly committed to some activity is less likely to terminate the activity than one who is
uncommitted
...
To do
this requires recognition that commitment can be attached to many different aspects of a survey, such as
the source or the sponsor, the researcher, the topic and issues being studied, and/or the research process
itself
...
The following hypotheses (untested) can be proposed:
1
...

2
...
g
...
) is perceived as legitimate
...
The more behavior productive of disorder is perceived as legitimate, the less favorable the
attitude toward the survey
...
This shift to online research is due largely to reduced
cost, the availability of dynamic surveys using advanced survey flow logic, the ability to display
visually interesting and even interactive graphics, the ease of survey creation and administration,
and the ability to eliminate errors associated with data entry, coding and transcription
...

72
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Coverage Error
Coverage error occurs when the sample frame (the group from which the sample is
drawn) does not represent the population as a whole
...
In more traditional
research methods such as mail or telephone methodologies, samples are drawn from sources such
as telephone directories, driver’s license records, rolls of property owners, credit reports, and so
forth
...

E-mail list brokers offer panels and e-mail address lists that may be targeted to reduce
coverage error
...

These lists are typically double opt-in, meaning that the users have specifically indicated their
agreement to receive surveys or other promotional materials
...
Targeted specialty lists such as physicians of a given specialty are
expensive, costing as much as $100 per completed response
...
E-mail name brokers make a practice of
not providing the list, but of sending the survey invitation out, thereby controlling their list and
avoiding survey abuse of the potential respondents on the list
...
Therefore
coverage error will continue to be the greatest source of inaccuracy for online surveys for many
years to come
...
Carefully selecting lists
from well constructed probability panels or panels having millions of members will help to
reduce coverage error
...
The estimation of sampling error requires that probability sampling methods be used,
where every element of the frame population has a known nonzero probability of being selected,
which may be made the same (i
...
, equal) for all
...
In online surveys the degree
of sampling error is generally unknown unless the sample is drawn from an online panel or other
frame with known size and characteristics
...

Online surveys are therefore subject to certain amounts of sampling error
...
This is an easy task, especially when using
an online panel
...
Spam filters, like caller
ID monitoring, prevent many survey requests from reaching the “In Box
...
This
73
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

self-selection bias is manifest as potential respondents consider the appeal of the survey topic,
survey length, and incentives to complete the survey
...

Increasing response rates and reducing nonresponse error most often includes the use of
multiple notifications and requests, and the use of personalization in the contact email requesting
completion of the interview
...
When a population of interest is not adequately represented online or is particularly
difficult to interview, a mixed-mode survey strategy should be considered to reduce nonresponse
error: a combination of e-mail and telephone, mail, or mall intercept techniques
...
She requested that the traveler complete the
airline satisfaction survey when they returned home or to their office
...
Most travelers kindly accepted the card
...
While many studies have confirmed
this fact, one of the more rigorous studies compared response rates for mail and e-mail surveys
(Shaffer and Dillman, 1998)
...
Results showed no statistically significant
difference between the 57
...
0 percent
response rate for the e-mail group
...
5 Increasing Response Rates
The variation in response rates for surveys is enormous, especially when interest and incentives are
considered
...
com, relates his experience with
three client surveys that differed greatly in their respective response rates (Smith, 2007)
...
The first survey consisted of a short 10-question survey entitled “What Do Women Want
...
Recipients of the survey were offered the chance to win $500 cash in a random drawing
and in addition were promised a copy of the results
...

2
...
This survey measuring attitudes and behaviors was sent through the same broker to a
random sample of “Internet shoppers
...
The university sponsorship of the survey was identified
in the cover letter that contained the professor’s name, contact information, and link to the
survey
...

A parallel paper and pencil survey was conducted for comparison purposes using a national
sample provided by Experian, a provider of credit rating reports
...
Smith and Gerald S
...
The mail version produced a 20 percent response rate
...
This
difference was attributed to the difference in sampling frames
...
However,
demographic differences aside, the psychographic profiles related to online shopping were
nearly identical
...
A third survey sent to brokers by a leading investment firm resulted in a
...
Further follow up revealed that this fast paced group of potential respondents
was too busy to be bothered with a survey
...
Perhaps most noteworthy is the finding
that when compared with the mail survey, the survey administered by e-mail produced 12
...
Individual item
response rates and item completion rates were also higher
...

Non-response is also important at the question level
...

Albaum et
...
(2010) show that contrary to expectations, using forced response increases both
the total response rate and the data quality improve
...
This applies not
only in traditional mail surveys but also in e-mail surveys
...
5 great
response rate variations can exist
...
The integration of panel information with the survey (using embedded
codes and data) facilitates the identification and tracking of survey respondents and
nonrespondents
...

Additional technologies enable the careful tracking of respondents
...
While technological advances help the
researcher to reduce non-response rates, it is clear that multiple factors are responsible for
75
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

nonresponse rates, many of which are not addressable through the administration and handling of
the survey
...
Measurement error may be due to such factors as faulty wording of questions, poor
preparation of graphical images, respondent misinterpretation of the question, or incorrect
answers provided by the respondent
...
For telephone and personal interviews,
measurement error will often occur when the interviewer misinterprets responses, makes errors
recording responses, or makes incorrect inferences in reporting the data
...
The size and
resolution of the monitor, browser, operating system (Mac, Microsoft Windows, Linux), and
even web page color pallet may change the appearance of the survey
...

Researchers have for decades compared measurement error differences for the various
modes of data collection
...
Standard surveys follow the structured/direct
approach for data collection and change little when transitioning from paper and pencil to radio
button or checkbox formats
...

Many of the traditional measurement errors associated with transcription and recording of
data are eliminated with electronic real-time entry of the data
...
In one recent survey of programmers and software developers conducted by Qualtrics
...
com gift
certificate
...

Online studies may be completed in 24 hours, as compared to the four to ten weeks
required for paper-and-pencil methodologies
...
These many steps involve many participants with
varying levels of expertise, and each may introduce error
...

Efficiencies aside, no matter which mode is used for survey completion, error control must be
addressed to assure quality results
...
Smith and Gerald S
...
It then considered communication as a means to obtain information from
respondents
...

The next section introduced the concepts of inaccuracy and ambiguity as the major
sources of response and non-response bias
...
Methods of
reducing non-response error were then discussed in the context of theories of survey response
...

The objective of marketing research is to understand the consumer and apply information
and knowledge for mutual benefit
...
In this chapter we have stressed the need to improve the
research process by reducing errors in the research process
...
New technologies continue to be developed, but each must be tested
for applicability under specific research conditions, and refined so that marketers are able to
better identify and measure the constructs being investigated
...
Roster, James Wiley, John Rossiter, and Scott M
...

Designing Web Surveys in Marketing Research: Does Use of Forced Answering Affect
Completion Rates?, Working paper
...
, Albaum, G
...
(1999, April)
...
Journal of the Market Research Society, 41, 2, 227–244
Bradburn, N
...
(1979)
...
San
Francisco: Jossey-Bass
...
J
...
B
...
© 1985, William C
...
Reprinted with permission of The McGraw-Hill Companies, pp
...
M
...
(1986)
...
Beverly Hills, CA: Sage
...
J
...
Reducing interviewer-related error through interviewer training,
supervision, and other means
...
P
...
M
...
E
...
A
...
Sudman, Measurement errors in surveys (pp
...
New York: Wiley
...
J
...
W
...
Standardized survey interviewing: Minimizing
interviewer-related errors
...
Smith and Gerald S
...
, Scott M
...
Geurts (2009), “Improving Survey
Completion Rates and Sample Representativeness Using Highly-Interesting Questions: A
National Panel Experiment Comparing One and Two Stage Questions”
http://papers
...
com/sol3/papers
...
G
...
It seems like only yesterday: The nature and consequences of telescoping
errors in marketing research
...

Semon, T
...
(2000a, August 14)
...
Marketing News, 34,
10
...
T
...
If you think a question is stupid—it is
...

Schaffer, D
...
, & Dillman, D
...
(1998)
...
Public Opinion Quarterly, 62, 378–397
...
(2007)
...

Stewart, C
...
and Cash, W
...
, Interviewing Principles and Practices, 4/e
...
William C
...
Reprinted with permission of The McGraw-Hill Companies, pp
...
, Finn, A
...
(1984, Summer)
...
Public Opinion Quarterly, 48, 520–524
...
D
...
E
...
(2003)
...
Public Opinion Quarterly, 67, 126–138
...
Smith and Gerald S
...
Respondents may be interviewed in person, by telephone, or they may
be sent a questionnaire
...


PERSONAL INTERVIEWS
As the name implies, the personal interview consists of an interviewer asking questions of one or
more respondents in a face-to-face situation
...
Recording of the
information may be done either during or after the interview
...

While it is significantly more expensive on a per-completed-interview basis, the personal
interview, as a collection medium, has several advantages relative to telephone interviews and
mail questionnaires
...
It also gives the opportunity
to obtain more information, as a personal interview can be of substantially greater length than
either a telephone interview or mail questionnaire
...
More
freedom is provided for adapting and interpreting questions as the situation requires, especially
in the case of unstructured personal interviews where visual, auditory, or olfactory aids are used
...
Problems with personal
interviews arise from its very nature in that it is a social interaction between strangers, often on
the respondent’s territory, initiated by an interviewer who may have little in common with the
respondent
...
This method avoids the logistic,
financial and time costs of travel to meet with respondents
...
Presumably the people are chosen on the basis of a predetermined sampling plan
...

The mall-intercept method is a widely used method of data collection in marketing
research
...
These facilities may be equipped with videotape equipment, private interviewing
compartments, food preparation facilities for taste tests, and a variety of other research
equipment
...
Researchers can observe an interviewer’s
technique and check the completed work immediately
...
Smith and Gerald S
...

EXHIBIT 5
...
They allow researchers to conduct visual, auditory and taste tests of ads, products and other physical stimuli
...
They offer an opportunity to obtain immediate response
...
They potentially provide more depth of response than non-face-to-face interviews
...
Researchers can use equipment to analyze responses (for example, voice-pitch or eye movement tracking
analysis)
...
A large number of respondents from a wide geographic area can be interviewed in a limited time
...
Researchers can control the interviewing environment and supervise the interviewer
...
However, it is becoming increasingly more difficult to locate people at home, and
even more people are hesitant to let strangers inside
...
For studying certain types of products or behaviors, the mall is a
more realistic setting when a respondent is being asked to make choices
...

Despite all these virtues, mall intercepts have limitations:
1
...

2
...

3
...

4
...


A variation of the mall intercept that is often used in business is to interview at
conferences, sales meetings, or other gatherings representing the population of interest
...
Respondents are most likely to recall and discuss their experiences
during the experience, not days later during a survey
...
Stream of consciousness interview
...

2
...
This asks for spontaneous, minimally prompted
reactions of customers to their environment
...
Directed general-response interview
...

4
...
This is useful to determine why consumers feel
as they do, as indicated by answers to other questions
...
Smith and Gerald S
...
Prompted reaction to execution elements
...
For example, an in-store taste test might include the question “What
do you think about the taste of China Sea brand Spring Rolls?”
The obvious advantages of on-site interviews are that the respondent is usually in a
proper state of mind and has better task or product recall
...
On-site interviews seem to produce
more robust information
...

But, do not seem to influence the quality of the data collected (Davern, Rockwood, Sherrod, and
Campbell, 2003)
...
Compared to e-mail or mail
surveys, telephone interviews often are more costly in terms of total costs of data collection
...
In addition, telephone surveys offer the
opportunity to probe for clarification or further information
...

In addition, when a telephone survey is conducted from a call center, they can be better
supervised than personal interviews
...
However, when the
population to be studied is business decision makers, some research practitioners believe that
more information may be obtained using the telephone than by conducting focus groups
(Eisenfeld, 2003)
...

With a detailed database to use as a sample frame, interviews of business people, current
customers, former customers, and prospects all can be contacted relatively easily
...
A recent study of political telephone surveys concluded
that advance pre-notification letters can significantly increase response rates (Goldstein &
Jennings, 2002)
...
Telephone surveys are unique in that they allow the
interviewer to respond to the potential respondent and attempt to turn a refusal into a completed
interview
...
These
responses can help the researcher handle objections and refine their interviewing skills
...
1
...
Smith and Gerald S
...
1
Possible Answers to Reasons for Refusals
Reasons for
Refusing
Possible Responses
Too busy
Bad health

Too old

Feel inadequate:
Don’t know
enough to answer

Not interested
No one else’s
business what I
think
Objects to surveys

Objects to
telephone surveys

This should only take a few minutes
...
When would be a good time for me to call in the next day or two?
I’m sorry to hear that
...
Would that be okay? (If lengthy or serious illness, substitute another family member
...
)
Older people’s opinions are just as important in this particular survey as anyone else’s
...
We really do want your opinion
...
They mostly concern your attitudes about local
recreation areas and activities, rather than how much you know about certain things
...
Maybe I
could read just a few questions to you and you can see what they are like
...
So, I’d really like to talk with you
...
Protecting
people’s privacy is one of our major concerns, and to do it people’s names are separated
from the answers just as soon as the interview is over
...

We think this particular survey is very important because the questions are ones that
people in parks and recreation want to know answers to, so they would really like to have
your opinion
...


SOURCE: Reprinted from Mail and Telephone Surveys: The Total Design Method by Dillman, D
...
This material is used by permission of John Wiley & Sons, Inc
...

Incentives may be promised or sent in advance with a preliminary letter when the mailing
address of the potential respondent is known, or they may also be offered when the initial request
for participation is a refusal
...

The main purpose of such incentives is to generate a greater response rate with the effect
of reducing nonresponse error
...
First, total cost
will increase, although cost per response may decrease depending on how effective the incentive
is
...
Third, sample composition may be affected, again with a positive or
negative effect
...
Finally,
interviewer effort may be affected
...
It is obvious that there must be sufficiently high telephone penetration in the segment
for this mode of data collection to be advantageous
...

82
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The basic limitations of telephone interviews are the relatively limited amounts of
information that can be obtained (at least compared with alternative methods) and the bias that
exists in any sample of home telephone subscribers
...
busreslab
...
htm)
...

Another additional problem for telephone researchers is that home telephone subscribers
are disappearing
...

http://www
...
com/business/articles/2009/07/20/20090720biz-cellnation0721side
...
Often this is manifested by inadequate efforts to complete
interviews with some of the harder-to-reach respondents
...

Another aspect of interviewer performance that can influence response rates and data
quality is actually something beyond any given interviewer’s control
...
There is evidence that accent can influence respondents’ participation
...
In the United States
there are many region-specific accents (New England, the Deep South) and also those that are
cultural (Hispanic, Asian, Indian)
...
When communication becomes difficult, refusals will increase
...
In addition, respondents who identify an interviewer’s accent may apply preconceived
biases to the interviewer and to the survey
...
At the very least, if a study is regional in nature, then having
interviewers from that region will also reduce nonresponse and bias
...

One way people use answering machines is to screen incoming calls
...
One
might be tempted to assume that screening and leaving messages would allow potential
respondents to choose to not participate—a form of refusal
...
This suggests
that where answering machines are operating, the call can represent a form of pre-notification
...
But
this phenomenon also has the potential to generate bias in the sample (Oldendick & Link, 1994)
...

This was the creation of the National Do Not Call Registry
...
Of primary concern to the research community is the use by telemarketers of
selling-under-the-guise-of-research techniques
...
Smith and Gerald S
...
A good overview of
telephone surveys is given by Bourque and Fielder (2003a)
...
Mail questions provide great
versatility at relatively low cost and are particularly cost effective when included as part of a
scheduled mailing, such as a monthly correspondence or billing
...
Respondents
remain anonymous unless a name is requested, the questionnaire is openly coded, or some
ethically-questionable practice is employed
...
If the time given is reasonable, say
one or two weeks, stating a deadline should not adversely affect the response rate
...

The overall process of data collection from a mail survey is summarized below as a
sequence of contact activities for an optimal mail survey
...
When designing a
survey, the researcher must consider issues that can affect response rate and data quality,
including those shown in Table 5
...

TABLE 5
...
Typically,
people indifferent to the topic being researched will not respond
...
e
...
But even with added mailings,
response to mail questionnaires is generally a small percentage of those sent; the modal response
rate is often only 20 to 40 percent
...
Smith and Gerald S
...

Other experimental research has evaluated response effects of questionnaire format and
length, survey sponsorship, endorsement, type of postage, personalization, type of cover letter,
anonymity and confidentiality, deadline date premiums and rewards, perceived time for task, and
the use of a follow-up reminder,
...

Cover letters are included as the first page of a questionnaire is shown in Exhibit 5
...

This letter could have been a separate page if more space was needed for the questionnaire
...
There
is no evidence that any alternative is universally better than another within each dimension
...
Further discussion of these will be found in the many
review articles and studies published in such sources as the Journal of Marketing Research and
Public Opinion Quarterly, in the book by Bourque and Fielder (2003b), and in the classic works
of Dillman (1978, 2000)
...
2 Example of a Cover Letter
My colleague, Dr
...
The broad
objective of this study is to gain an understanding of how people use banks and similar financial organizations, and
what characteristics influence their behavior with such companies
...

The enclosed questionnaire is being sent to a large number of the customers of the Bank of Anytown, each of
whom has been selected by a random process
...
Please note that you do not have to add postage to this envelope
...
No person associated with The Bank of Anytown will see
any questionnaire
...
No person other than Dr
...
If you do not wish to participate
in this survey simply discard the questionnaire
...

The code number at the top of the questionnaire will be used only for identifying those people who have not
responded so that he or she will not be burdened by receiving a follow-up mailing
...

The success of this project depends upon the assistance of persons such as yourself
...

Sincerely,
Gerald Albaum
Professor of Marketing

85
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Endorsements are an intriguing dimension of survey research
...
” An endorsement can be included as a company logo or a person under who’s
signature the letter is sent
...

Rochford and Venable (1995) found that significantly higher response rates were
observed when there was endorsement by an external party associated with the targeted audience
than when there was no such endorsement
...

Since people responding to a mail questionnaire tend to do so because they have stronger
feelings about the subject than the nonrespondents, biased results are to be expected
...
This is a type of nonresponse validation
...
The initial low cost per mailing may
therefore be illusory
...

Variations on Mail Interviews

Many variations of the mail interview are frequently used
...

Warranty cards often ask for information about where the item was purchased, what kind
of store or outlet sold it, when it was purchased, and other variables such as demographics and
life style
...

Another variation is the questionnaire that is either printed or inserted in a newspaper or
magazine
...
Among the many problems with this approach is the lack of any formalized
control over the sample
...

Another modification of including a questionnaire in a printed publication asks
respondents to return the questionnaire by fax
...
But, again there is little control over the
sample
...

Potential advantages of using fax include quick contact, rapid response, retention of the
original document format and visual images used, modest cost, and automated faxing that works
directly with the computer (Baker, Hozier, & Rogers, 1999)
...
3
...
The results were then available online in real time
...
Smith and Gerald S
...
Web and e-mail surveys are fulfilling their promise to be a driving force in marketing
research
...
S
...

http://www
...
com/interactive/home-internet-access-in-us-still-room-for-growth-8280/

The Internet has experienced a growth rate that has exceeded any other modern
technology, including the telephone, VCR, and even TV
...
At the opposite end of the
spectrum, the approximately 25% who are non-adopters include a disproportionate number of
elderly, single mothers, African Americans and Hispanics, and lower-income individuals
...
However even today, most groups like company
employees, students, or association members have nearly 100 percent Internet access and check
e-mail on a daily basis
...
1

Example of Questionnaire Processed by Fax

87
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Electronics driven lifestyles that include online social networks, massive use of texting
and pervasive internet connections are no doubt responsible in part for seemingly responsible for
attitude and behavioral changes in the way we view our increasingly virtual world
...
These online customers provide
excellent access for research purposes
...
As a result, online
marketing research has become so widely accepted that online research has been optimistically
projected to account for as much as half of all marketing research revenue, topping $3 billion
...
This pace will
continue to increase as new modalities for research open: wireless PDAs, Internet-capable
mobile phones, Internet TVs, and other Internet-based appliances yet to be announced
...

Substantial benefits accrue from the various approaches to computer-assisted data
collection in surveys as shown below:
Respondents need few computer related skills
Respondent choose their own schedule for completing survey
Can easily incorporate complex branching into survey
Can easily pipe and use respondent generated words in questions throughout the survey
Can accurately measure response times of respondents to key questions
Can easily display a variety of graphics and directly relate them to questions
Eliminates need to encode data from paper surveys
Errors in data less likely, compared to equivalent manual method
Speedier data collection and encoding compared to equivalent manual method
...
Researchers noted high correlations in aggregate purchase
intentions
...
Wilkie further reported
that while responses gathered using different survey modes may be similar, the demographic
profiles of online and traditional respondents groups do differ
...
However, the possibility of weighting actually
raises the question of whether it is better to model phone or mall intercept behavior (which are
also inaccurate) or to attempt to independently model the actual behavior of the respondents
...
Smith and Gerald S
...
3

Comparative Evaluation of Alternative Survey Methods of Data Collection
MallIntercept
Interviews

High

High

CAPI
Moderate
to high

Low

High

Low
Moderate
to high

High
Moderate
to high
Potentially
high

Moderate

Social desirability
Obtaining
sensitive
information
Potential for
interviewer bias

Mail
Surveys

Mail
Panels

Low

Low

High

Moderate

Moderate

Internet/
Web
Moderate
to high
Moderate
to high

High

High

Moderate

Moderate

Moderate

Low

Moderate
Moderate
to high

Moderate
Low to
moderate

Moderate
to high

High

High

Low

Low

Low

Moderate
Low
Moderate

Low
High
High

Moderate
Moderate
High

Moderate
Moderate
High

High
Moderate
Low

High
High
Moderate

High
Moderate
Very low

Moderate

Low

Low

High

High

High

Moderate

High

High

Low
Moderate
to high

Low

Low

Low

High

Low

Low

Low to
moderate

High

Moderate
to high

High

Moderate

High
Moderate

Moderate

Sample Control
Control of data
collection
environment
Control of field
force
Quantity of data
Response rate
Perceived
anonymity of
respondent

Cost

In-Home
Interviews

High

Criteria
Flexibility of data
collection
Diversity of
questions
Use of physical
stimuli

Speed

Telephone
CATI
Moderate
to high

High

High
Moderate
to high
Moderate
to high

Low
Moderate
to high
Moderate
to high

None
Low
Low

None
Low to
moderate
Low to
moderate

None
Very high
Low

SOURCE: From Malhotra, N
...

Reprinted with permission of Pearson Education, Inc
...


PROBABILITY AND NONPROBABILITY SURVEY APPROACHES
A variety of approaches to presentation of surveys and recruitment of respondents are
used on the Web
...
However the
majority of online research can be typified as a one-shot mailout, the objective of which is to
obtain a sufficient number of completed responses
...
The question is, then, what is required for effective online research?
Online Nonprobability Surveys
Nonprobability samples offer neither of these features
...

As an example of the Web site interest variety of survey, Figure 5
...
com
“Quick Vote” survey, which includes a link to the online results page
...
Smith and Gerald S
...
2 Web Site Interest Survey

Source: http://edition
...
com/

The National Geographic Society Web site offers surveys that focus on a variety of
educational, social and environmental issues at www
...
com/geosurvey/
...
The 2006 Global Geographic Literacy Survey was conducted jointly with Roper
Research to assess the geographic knowledge of people ages 18 to 24 across the United States
...

Other well-recognized nonprobability surveys include the ACNielsen BASES (test
marketing) and Harris-Black panels
...
The parallel telephone and mall intercept studies provide weighting to proportionately
adjust online samples to reduce selection bias
...
Coverage
errors, nonresponse errors, and measurement errors still apply and may reduce the
generalizability of the data
...
When the target population is
large, random samples from the list will be used
...

Where the target population of interest is visitors to a given Web site, pop-up surveys
may be presented randomly to visitors during their visit to the site
...
The
Qualtrics
...
Surveys, white papers, and messages
can be distributed based on a variety of conditions, including multiple pages visited that contain
specific key words
...

90
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Pre-recruited online panels, when recruited using probability-based sampling methods
such as random-digit telephone dialing, also produce probability surveys
...
The initial interview may also include a
variety of demographic, lifestyle, and computer usage questions that would help in weighting the
panel, thereby reducing selection bias
...
Respondents
contacted by mall intercept, telephone, mail, or other probability-based sampling mechanism, are
given the opportunity to respond in several modes, including online
...

Wisconsin cheese producers respond annually to an industry group survey that reports
production by the type of cheese
...
When asked if they would prefer a paperand-pencil or online survey, more than 50 percent favored the online mode
...
A mixed-mode survey design is the obvious choice
...

These factors include nonsampling errors unique to the Internet methodology: for example,
fewer households have adopted the Internet than have telephone or mail and researchers lack
control of the respondent’s computer configuration (browser, operating system, fonts, and
resolution)
...
(Sources of nonsampling
error were discussed in Chapter 2
...
Marketing researchers, both professional and casual, often neglect to consider
the implications that nonprobability sampling and surveys have on the ability to make inferences
regarding the target population
...
In Chapter 10, we build upon this discussion of general survey and sampling
methodology
...
The total package concept
underlying survey design strategy goes beyond contacts with respondents
...
This is the essence of a total design
...

Pretesting and Pilot Survey
A distinction should be made between a pretest and a pilot survey
...
Smith and Gerald S
...
In contrast, a pilot survey is a small-scale test of what the survey is to be,
including all activities that will go into the final survey
...
Are we asking “good” questions?
2
...
A general rule of thumb for most surveys is that a pretest of
about 30 to100 interviews is adequate, provided this covers all subgroups in the main survey
population
...

The pilot study is designed to ascertain whether all the elements in the survey fit together
...
One
aspect of the pilot survey is that it can help researchers decide the size of the original sample for
the main survey
...

Both pretesting and pilot surveys can provide information helpful to manage some of the
sources of potential research error
...


LONGITUDINAL DATA COLLECTION WITH PANELS
Panels are widely used in marketing research
...
In this chapter we have discussed methods for
reducing response error along with some general characteristics of panels
...

For example, panels have been effectively used to develop early forecasts of long-run
sales of new products
...
In addition, several universities
maintain consumer panels to obtain research data and generate revenues by providing
data to others
...

The distinguishing feature of a panel is the ability to repeat data collection and collect
longitudinal data from a sample of respondents
...
Panel studies offer at least three advantages over onetime surveys:
1
...

2
...

92
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

3
...

When responses are obtained at two or more times, the researcher assumes that an event
happens or can happen (i
...
, changes may occur) during the time interval of interest
...

Assume that we have changed the package in one market for a brand of paper towels
called Wipe, and that we run a survey of 200 people purchasing the product two weeks
before the change (T1) and a similar measure for the week after (T2)
...
4
...
However, only the turnover
table from the panel in (B) can tell us that 20 former buyers of Wipe switched to X and
that 40 former buyers of X switched to Wipe
...

TABLE 5
...
This
consciousness of continuing participation can lead to panel conditioning, which may bias
responses relative to what would be obtained through a cross-sectional study
...
Another issue of concern is panel attrition, the extent of
nonresponse that occurs in later waves of study interviewing
...

There are many distinguishing characteristic of panel types
...
Panels can also be characterized by geographic
coverage (ranging from national to local), whether a diary is used, data collection method (all
types are used), sampling method employed for a given study (probability or not), and type of
respondent
...
Smith and Gerald S
...
This panel involves recruiting
shoppers in the target market area to participate in the panel, and each person typically is
compensated for participation
...
At the end of a normal shopping trip in a cooperating store, the
card is presented at the start of the checkout process
...
The types of information available from this sort of panel
are similar to those discussed in the preceding chapter for scanner-based syndicated services
...

One last comment about panels is that they are often used for a cross-sectional study
...


SUMMARY
This chapter first examined the media through which interviews may be conducted
...
Variations of these basic methods,
including electronic-based variations, were briefly described
...

The objective of marketing research is to understand the consumer and apply information
and knowledge for mutual benefit
...
In this chapter we have stressed the advantages as well as
the caveats associated with online research
...
As new technologies continue to be developed, they are
tested for applicability in marketing research settings, and refined so that marketers are able to
better identify the needs and wants of today’s consumers
...
Smith and Gerald S
...
azcentral
...
html
Baker, K
...
, Hozier, G
...
, Jr
...
D
...
E-Mail, mail and fax survey
research: Response pattern comparisons
...

Becker, E
...
(1995, January 2)
...
Marketing News, 29, 9
...
B
...
P
...
How to conduct telephone surveys
...

Bourque, L
...
, & Fielder, E
...
(2003b)
...

Thousand Oaks, CA: Sage
...
busreslab
...
htm
www
...
com/
Davern, M
...
H
...
, & Campbell, S
...
Prepaid monetary incentives
and data quality in face-to-face interviews
...

Dillman, D
...
(1978)
...
New York: WileyInterscience
...
A
...
Mail and internet surveys: The tailored design method (2nd ed
...

Eisenfeld, B
...
Phone interviews may garner more data
...

European Society for Opinion and Marketing Research
...
Distinguishing telephone
research from telemarketing: ESOMAR guidelines
...

Goldstein, K
...
, & Jennings, M
...
(2002)
...
Public Opinion Quarterly, 66, 608–617
...
(1997, January 6)
...
Marketing News, 31,
14
...
K
...
Marketing research: An applied orientation (4th ed
...

www
...
com/interactive/home-internet-access-in-us-still-room-for-growth-8280/
Miller, T
...
(2001, September 24)
...
Marketing
News, 30
...
Smith and Gerald S
...
nationalgeographic
...
W
...
The answering machine generation
...

Rochford, L
...
F
...
Surveying a targeted population segment: The
effects of endorsement on mail questionnaire response rate
...
86–97
...
T
...
Most research firms use mall intercepts
...

Wiseman, F
...
, & Schafer, R
...
An experimental test of the effects of a
monetary incentive on cooperation rates and data collection costs in central-location
interviewing
...

Xu, M
...
J
...
C
...
The impact of messages on survey participation
in answering machine households
...


96
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 6
QUALITATIVE RESEARCH AND OBSERVATION
In Chapter 4 we discussed communication as a means of obtaining information from
respondents
...
It will be recalled that at the extremes, an interview will be
either structured or unstructured in organization and that it may be either direct or indirect in
questioning
...
We conclude the chapter with an assessment of direct and indirect research
techniques
...
As we mentioned in Chapter 2, a focus group interview involves group of
people jointly participating in an interview that does not use a structured question and- answer
method to obtain information from these people
...

The composition of the group varies according to the needs of the client, especially the problem
under study
...
Such applications include gaining greater understanding of consumer
wants, thoughts and needs; idea generation (e
...
, problem, unmet needs, ideas for new products),
concept development and screening, tests for comprehension of promotion and communication
materials, and establishment of “opinion leader” panels
...
Practitioner researchers accept the idea that qualitative techniques like focus
groups, explore, define and describe
...
These techniques are complements, not substitutes
...

Example: The Federal Duck Stamp Office of the Department of Interior, U
...
Fish and Wildlife
Service (Rydholm, 2000) focus group study
...
They are also popular
among stamp collectors
...
Because the number of hunters has been declining,
there was a need to find a way to reach a new audience and to broaden the market for the stamps
...
The objective was to appeal to bird watchers, hikers, and
other naturalists
...

The certificate idea came directly from the focus group research
...
Once a range of appeals
97
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

was identified in the groups, the questioning centered on isolating the elements that made people
change their minds
...


Raymond Johnson (1988) has identified four distinctive categories of focus groups on the
basis of examining tapes from the project files of several research companies
...
The focus group types are as follows:
Exploratory studies of consumer lifestyles and probing to “just find out what’s on the
consumers’ minds these days
...
Potential users are able to react to a concept
still in its formative or experimental stage
...
The topic is
framed by the moderator’s instructions to describe, usually by situation-specific
narratives, the details of personal experiences in using a particular product or service
...
All types of media may be covered
...

The use of focus groups is not limited to consumer products and services
...
The ways in which focus groups
are structured and conducted are similar for consumer-based and business-to-business groups,
except as identified by Fedder (1990)
...
This means that a sense of what is unique
about customers is gained —their self-perceptions, desires, and needs that affect everything they
do
...
Qualitative research offers not just an intellectual comprehension of consumers but a
vivid, visceral recognition that affects, on a very deep level, how clients see, feel about, and deal
with their customers from then on
...
1
...
2)
...
To accomplish this,
the moderator needs to explain to the group the operational procedures to be used, the group’s
purpose, and why the group is being recorded
...
The respondents need to feel relaxed early on, and often moderator
humor helps this to happen
...
A moderator has
98
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

done a good job if, after the topic has been introduced, the discussion proceeds with little
direction from the moderator
...
Thus, a
critical measure of success is that the group members have talked with each other, not with the
moderator
...
1 Observation of Focus Groups by Clients
As an observer, a client should be cognizant of certain things as he or she observes the dynamics of a
focus group in the “back room” from behind a one-way mirror
...


Determine your overall impression of the people in the group by looking at their sophistication level,
appearance, and the way they express themselves
...


Do the respondents’ reactions support your assumptions? A way to assess whether the people are
atypical or less than fully honest is to have outside data
...


Are there segments that seem to exist in the focus groups, perhaps based on psychographics?

4
...
Look for variance! Do not count numbers
...


A single comment by a respondent may be quite insightful
...


Look at how people say things, not just what they say
...


7
...


Although certain statements by respondents may appear to be complimentary, they may not really
be
...


9
...


10
...
After the session, talk with the moderator and ask him or her to put the responses in perspective
...
Never take what is said as being personal
...

12
...
Seeing and hearing

consumers, up close, has an impact that no set of data and no written report alone can have
...


When conducting focus groups with professionals, the practical aspects of focus group
use are somewhat unique
...
When professionals are
involved, the moderator’s interaction must be such that he or she is presented as an authority on
research in the field of the professional
...

99
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

EXHIBIT 6
...
As the costs of doing focus groups increase,
some companies consider hiring an “outsider” to conduct groups
...

2
...

4
...

Company personnel are more familiar with the product and will be able to ask the right
questions, recognize the nuances, and know when the group is not telling the truth
...

There have been bad experiences with the outside moderators used in the past
...
However,
Thomas Greenbaum (1991) believes that when the total situation is considered, it is preferable to hire the
right outsider to conduct the focus groups
...

2
...

4
...

6
...


An outside moderator can be more objective than an insider, and is less likely to lead the group
in a specific direction
...

An outside moderator can be helpful in designing the groups and in developing an effective
moderator guide
...
Since
moderators pretend not to know much about the category being discussed, it is possible to ask
the seemingly dumb question or seek “help” from the participants that will generate information
to achieve the goals of the research
...

An outsider will be more objective in interpreting results than will an insider
...


It seems clear that much of the argument for using outsiders is based on their greater experience and
objectivity
...
Greenbaum (1999) presents a practical perspective and a
more detailed guide of moderators
...
Respondents are recruited from across the country and are
told to call a toll-free number at a certain time to participate
...
Simon (1988)
listed some of the advantages of this approach:
1
...

2
...

3
...

100
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

4
...

5
...

6
...

7
...

8
...

9
...

One disadvantage of the telephone focus group is that the researcher loses the ability to
evaluate verbal response within the context of nonverbal communication
...
One company has developed video software capable of broadcasting
live focus group interviews from a nationwide network of independently-owned focus group
facilities to a conference room in a company’s own office
...

They can maintain audio contact with the moderator, video the complete sessions, and hold
“open-mike” post-group debriefings
...
Such video-system focus-group facilities have expanded to include global options
...

Videoconferencing is being used to reduce costs
...
With the right convention, a large concentration of the
target market can be accessed (Greenbaum, 1993)
...
This so-called virtual research has the appeal of groups being run
more quickly and at lower cost
...
That is, body
language, voice inflection, facial expression, and interaction between people cannot be observed
(Miller, 1994)
...

Most of these techniques employ the principle of projection
...
The person giving the description will tend to interpret the situation in terms of his or
her own needs, motives, and values
...

Projection techniques include word association tasks, sentence completion tests, and
interpretation of pictures and pictorial representations that have been developed as a means of
inducing people to project their feelings (see Table 6
...
These techniques have been most
widely used for studies of consumer products that are similar in quality, performance, and
price—notably for such products as automobiles, soaps and detergents, gasoline, cigarettes, food
products, beverages, and drug sundries
...
Because projective techniques are designed to bypass
people’s built-in censoring mechanisms, they are useful in eliciting information about sensitive
or threatening topics and products
...
Smith and Gerald S
...
1 Classification of Projective Techniques
Technique

Response Requested of Subjects

Construction
Ordering
Item preference test
Expressive techniques

To create a story based on a stimuli presented
To order stimulus items by preference

Association
Word-association test
Completion
Sentence-completion test

Play a role; draw a picture of a person doing something;
describe a character in a simulated situation
To reply to a stimulus with the first word, image, or percept
that comes to mind
To complete incomplete expressions, images, or situations

Most indirect interviews are at least partially structured in that they are conducted using a
predefined set of words, statements, cartoons, pictures, or other representation to which the
subject is asked to respond
...
Indirect interviews, therefore,
are commonly neither fully structured nor fully unstructured; ordinarily they utilize both types of
question
...

The Third-Person Technique
The simplest way of obtaining information through indirect questioning of a respondent
is to ask for the view of a neighbor, an (unnamed) associate, or some other person whose views
on the subject at hand might reasonably be known
...

An early study using a variation of this technique that has come to be regarded as a
classic is the study by Mason Haire (1950) on instant coffee
...
The purpose of the study was to determine the
motivations of consumers toward instant coffee in general and Nescafe, a brand of instant coffee,
in particular
...
Among the questions asked were “Do you use instant coffee?” and (if “No”)
“What do you dislike about it?” The majority of the unfavorable responses were of the general
content “I don’t like the flavor
...
An indirect approach was therefore chosen
...
” These
shopping lists were shown alternately to a sample of 100 respondents, each being unaware of the
other list
...
Try to project yourself into the situation as far as possible until
you can more or less characterize the woman who bought the groceries
...
Wherever possible indicate what factors
influenced your judgment
...
The descriptions given were summarized as follows (Haire,
1950, p
...
Smith and Gerald S
...

Forty-eight percent of the people described the woman who bought Nescafe as failing
to plan household purchases and schedules well; 12 percent described the woman
who bought Maxwell House this way
...

Sixteen percent described the Nescafe woman as not a good wife; 0 percent described
the Maxwell House woman this way; 4 percent described the Nescafe woman as a
good wife; 16 percent described the Maxwell House woman as a good wife
...
The woman using the instant coffee was
characterized as being lazier, less well organized, more of a spendthrift, and not as good a wife
as the one using the conventional coffee
...

This study has been replicated a number of times
...
Nevertheless, the original study remains as a classic application of
the third-person technique of research
...
The
respondent, by answering quickly, presumably gives the word that he or she associates most
closely with the stimulus word
...

Sentence Completion Tests
Sentence completion tests are similar to word association tests, both in concept and in
use
...

Recognizing that people may react in more than one way to a sentence stem, participants are
asked to fill in the sentence several times rather than once
...

Sentence completion tests provide a top-of-mind association between the respondent and
the topic/product/subject being investigated
...

Example: A sentence completion was used in a study of automobile buying to probe the
motivations of automobile buyers and thereby provide a sounder basis for advertising
...
227–228)
...
Smith and Gerald S
...

Women’s responses:

...


...


...


...


Sentence stem: A car of your own
...
is a pleasant convenience
...
is fine to have
...
is nice to have
...
you take good care of it
...
I want to make darn sure it has a good coat of wax
...
check the engine
...
how soon can I start polishing it
...
I would take care of it
...
is a good thing
...
absolutely a necessity
...
For men a car was something for which they should be
protective and responsible
...

Men appeared to feel closer to their car and regarded it as more of a necessity than did women
...
It is used to explore the underlying predispositions, needs, desires, feelings, and
emotions of the consumer toward products and services
...

Insofar as obtaining information on motivations is concerned, the concept of “depth”
refers to the level at which underlying motivations are uncovered
...
Mariampolski (1988) created a useful list of different probes that can be used:
1
...

3
...

5
...

7
...

9
...

11
...

13
...

15
...

17
...


The silent probe (use eye contact and body language)
Request elaboration
Request definition
Request word association
Request context or situation of use, occurrence or activity
Shift the context or situation
Request a clarification
Request a comparison
Request classification or typology
Compare and contrast to a previous statement
Challenge veracity
Challenge completeness
Confrontational probe
Echo probe
Interpretive probe
Summary probe
Purposive misunderstanding
Playing naive
19
...
Smith and Gerald S
...
By following leads and cues provided by respondents, phrasing questions to
continue the flow and pattern of the conversation and to maintain the rapport established, the
competent interviewer can explore and probe the underlying motivations of the respondent
...
The depth interview is ideal for obtaining from consumers anecdotes
of times they used a product or service, and such “stories” provide the marketer with a good view
of what products mean to consumers
...
Telephone interviewing has proved to be effective in obtaining such
consumer stories
...
Thus, each has its place in the total set of methodologies from which
marketing researchers can draw
...
Depth interviews are useful in exploratory research to obtain background information to
support the development of a quantitative survey instrument
...
Depth interviews may be the sole research method used such as hard-to-reach groups, or
when focus groups are not feasible
...

3
...
For instance, in the case of a rollout of a new
product the marketer may not want respondents to be influenced by the views of others
...
This approach uses a rigorous in-depth interviewing procedure to
understand the deep meaning customers attach to product decisions
...

ZMET identifies the metaphors that shape decision making and behavior
...

The ZMET interview focuses on photos or drawings that subjects are asked to bring with
them, representing their feelings about the topic of discussion for example, donating to a charity
or university, or purchasing a specific product or brand)
...

To better understand this result, Olson, Waltersdorff and Forr (2008) provide, in Figure
6
...


105
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 6
...
2 shows examples of these deep metaphors that each of us in expressing thoughts
and emotions
...

Table 6
...

References to the positive (or negative); to things that give one pleasure; feeling good or bad;
References to enjoyment, fun, happiness, euphoria, well-being
...

Considering an intangible idea, concept as a thing, a physical object, an entity; e
...
; "I can't get my
breath"
References to equilibrium, stability, equalize or compensate; Including both sides; Images of
scales, teeter-totter, balance beam; References to reciprocity--give and take; References to
'stable" emotional states such as calm, relaxed, serene; Feeling 'right" with the world
...

References to having/getting the requisite knowledge, energy, tools, or materials to accomplish
some task; Having or getting help and assistance from others (we are a team, I need support);

106
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Container

References to being in (or out) of a place (house, room); References to keeping or storing;
References to "in" and "out;" Keeping things out as well as in; Being wrapped up
...


Motion or
Movement
Journey
Transformation

References to moving (flowing, traveling, running or walking); References to movement actions
(hurrying, getting going); Keep moving; Keep it going
...

References to changing from one state to another-physical or emotional; Becoming something or
someone else; References to evolving, maturing, growing;
References to the passage of time; Reference to past events or historical perspectives; Images of
clocks and watches; References to old memories, remembering past events
...
References to breeding, evolving, growing
...

References to knowing, understanding, learning
...
In contrast, the unknown, ignorance,
stupidity; Inability to comprehend and understand something
...
Flight

Knowing and
knowledge
Contrast or
Paradox
Personal
expression
The Ideal
System

References to looking at both/two sides; Comparing opposites; Juxtaposition of opposites;
References to paradox--being one thing and its opposite at same time;
Reference to things that express one's personal goals, values, points of view
...

Reference to the ideal object, situation, feeling
...
References to
perfection, the perfect one
...
dirty, impure, devilish,
malevolent
References to love, fondness; warm relationships with family and friends; Caring for or nurturing
others, animals, plants; Being a caring resource for others; taking responsibility for and supporting
and helping others

Masculine/
Feminine
Birth/Death

References to gender--male or female; Having masculine or feminine traits; Attributions of gender
to nonliving things, ideas, concepts
...


Quantify/
References to quantification; Keeping track; Counting; Yardsticks, scales, sextons; Computing
Measure
amounts and quantities
Holism/
References to being whole, entire, complete, not lacking anything, not having weaknesses that
Completeness
compromise any other part of the whole
...
Olson, Olson Zaltman Associates (www
...
com)
...
Many standard qualitative techniques neglect
empathy
...
Smith and Gerald S
...

2
...

4
...

6
...


The researcher needs to imagine himself/herself in the respondent’s situation and must
listen to the respondent fully
...

Ask open-ended, non-leading questions that start with how, what, and why
...

Challenge generalizations by asking for specific examples
...

Let the respondent reveal himself/herself through personal studies
...


MEANS-END ANALYSIS
Means-end analysis, also known as Laddering and Means-End Chain, is an in-depth, oneon-one interviewing technique that is directed at identifying the linkages people make between
product attributes (the means), the benefits derived from those attributes (the consequences), and
the values or end state (the ends) that underlie the desire to achieve the consequences
...

Laddering interviews, as an in-depth interviewing technique, employ structured dialogues
that identify the most important attributes, the benefits (consequences) derived from those
attributes and the values linked to the consequences in a given usage situation (Figure 6
...
The
interviewer moves up and down the means-end chain, identifying the hierarchical structure of the
product attribute / product benefit / personal benefit / personal value components and linkages
...

Figure 6
...
Smith and Gerald S
...
This laddering of the reasons that underlie
a decision provides a much deeper understanding of the consumer than does a traditional “ratings
survey” of product attributes, but does not penetrate as deeply as does the ZMET analysis
...
Exhibit 6
...
In practice,
several different methods may be used to capture a full range of meaningful distinctions between
a brand and its competitors
...
Other tools, such as a worksheet (Exhibit 6
...

EXHIBIT 6
...
The respondent gives positive and negative associations for the brand or product
category, along with reasons why the characteristic is viewed that way
...

Grouping similar brands
...
The primary reasons, most
important attributes, and most representative brands are identified and attributes and consequences are
laddered
...
The usage context for a brand or product can be described either as physical
occasions (place, time, people), or need state occasions (relaxing, rejuvenating, building relationships,
feeling powerful, reducing stress, and getting organized)
...

Preference, usage, similarity and dissimilarity differences
...
Similarity and dissimilarity
groupings also provide a direct method of distinguishing between brands
...

Timing of purchase or consumption
...

For example, a respondent might be asked to identify products used for relief of a stuffy nose into several
stages like onset, full-blown, and on-the-mend, or daytime and nighttime
...

Usage trends
...
For example, respondents may be asked, “Will this
brand be used more often, less often, or about the same as you have used it in the past?” Then, reasons
for increased, decreased, or unchanged usage are discussed
...
Product and brand substitution methods elicit the degree of similarity of
perceived attributes and consequences associated with usage
...
The respondent can first sample or be given a brand descriptions, followed by questions like: how

109
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

likely would you be to substitute (name of the new brand) for your current brand for this occasion—why
is that?
Alternative usage occasions
...
Questions might be phrased to ask: why would you consider using Brand A
for this occasion, or what is keeping you from using Brand A for this occasion now? Both positive reasons
why a brand fits a new occasion and negative reasons why it does not fit can be elicited and laddered
...

Using the following types of probing questions, the study identified nine critical
consumer attributes for a new card: no annual fee, status, low interest rate, added value features,
acceptance, credit limit, ability to carry a balance, location of the sponsoring bank, and
availability
...

How has that experience changed your life?
Do you seek out similar experiences today?
How would you describe a card that could help you achieve this?

These attributes were found to be linked to 12 benefits (consequences) that were
perceived as part of card usage: not feeling cheated, independence, convenience, dependability,
saving money, financial responsibility, freedom, establishing a credit history, power, security,
supporting the community (local card), and confidence
...
This analysis is actually mapped to
show the hierarchy of attribute relationships
...
3, four distinct sub-hierarchies are shown that focus respectively on the four
values
...

In this case, it is easy to imagine four separate commercials, each stressing a different set of
attributes and consequences with overall personal value appeals to “Providing for the Family,”
“Superior Quality of Life,” “Peace of Mind,” and “Heightened Self-Esteem
...

The advantage of the benefit chain is that it is relatively easy to administer without a
highly sophisticated interviewer
...

110
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 6
...


OTHER TECHNIQUES
Other variations of projective techniques have been used that are similar in nature but
different in form from those already described
...
The usual procedure is to present the interviewee with a pack of cards on which
brand names or other stimuli are printed
...
Following this, the respondent is asked to describe a way in which any two of the
familiar brands (stimuli) are like each other and different from the third
...
The response may be in the
form of a paired comparison, a ranking, or a numerical rating on a scale
...
Additional respondents are interviewed until no new dimensions are given
...

An example of the use of the repertory grid is a study done for a major producer of frozen
foods who was interested in introducing a new type of toaster item that could be made to taste
like a variety of freshly baked products
...
The stimuli were 20
products as shown in Table 6
...
The study resulted in a list of 22 attributes that were scaled as
111
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

bipolar scales (see Chapter 7) for further analysis, as shown in Table 6
...
The stimuli and
attributes can be set up as a grid whereby respondents can then rate each stimulus (i
...
, bakery
item) on each attribute
...
3
Stimulus

List of Food-Item Stimuli Used in Repertory-Grid Study
Stimulus

1
...
Buttered toast (white)
3
...
Jelly donut
5
...
Blueberry muffin and margarine
7
...
Toast and marmalade
9
...
Toast and margarine (white)

11
...
Danish pastry
13
...
Chocolate chip cookie
15
...
Coffee cake
17
...
Toasted pound cake
19
...
Bagel and cream cheese

SOURCE: Green, Tull, and Albaum, 1988, p
...


Table 6
...
713
...
Smith and Gerald S
...
4 discusses the use of protocols (Ericsson & Simon, 1984)
...

EXHIBIT 6
...
This record is obtained by asking the respondent to “think out loud” or talk about
anything going through his or her head while performing the task
...

Protocols can be recorded concurrent with the task or retrospective when the verbalizing aloud is done
just after the task has been finished
...
The form and particular
stimuli to which the research subject should respond is not defined or specified by the researcher
...

Example: Comcast, a major provider of cable TV, telephone and Internet services offered an unsuccessful
self install kit that often resulted in customers calling for a technician to complete the installation
...
Comcast identified the key problem areas in
the install process and was able to revise installation instructions and provide an 800 support line, thereby
making significant reduction in install service calls
...

The general underlying principle is that the person will project his or her own psychological
interpretation of the situation into the response
...
When I mentioned XYZ, my partner remembered the last visit there
...
”

OBSERVATION
The remaining major method of collecting qualitative information is through observation
...
Rather
than asking respondents about their current behavior, it is often less costly and/or more accurate
if the behavior is observed
...
This exploratory qualitative methodology traces the stories of people, circumstances,
decisions and events leading to actual purchase decisions through one-on-one interviews
(Berstell & Nitterhouse, 2001)
...
Some key
characteristics of this approach are (Berstell, 1992):
Case studies uncover motivations through demonstrated actions, not through opinions
...

113
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Observation and documentation are used to stimulate questions and corroborate
responses
...

Case studies require researchers who are essentially “market detectives” rather than
“census takers
...

Observation may be used as the sole means of collecting data or, as is frequently the case,
it may be used in conjunction with other means
...
In some circumstances, observation is the only means of collecting the data desired
...
Brands usage
reports of well-established brands generally show a “halo effect,” an upward bias reflecting the
prestige the respondent associates with the use of the brand
...

Other examples include a food retailer who tested a new type of shelving for canned
goods by observing shoppers as they used the new shelves, and a toy manufacturer who, in a
laboratory setting, observes children playing with product prototypes
...

There are some problems in using observation
...

Since human perception is quite selective, what people observe depends upon their backgrounds
...
The second potential problem is that the behavior being observed is not
representative
...

That is, the behavior observed may be a unique incident
...
If a person is being observed, his or her behavior
may not be a “true” behavior
...
g
...
Thus, the situation and setting are critical to the observation
experience
...
The distributor audit is the more widely known of the two
...
As indicated there, data from this and audits available through other
research agencies provide estimates of market size, market share, geographic pattern of the
market, seasonal purchasing patterns, and results of promotional and pricing changes
...
In this type of audit, the field worker takes an inventory of the brands, quantities,
and package sizes that the consumer has on hand
...
When the audit is combined with questioning of the
consumer, an estimate of usage may be made
...
Its use has declined
as the use of consumer panels has increased
...
Smith and Gerald S
...
Some of these devices are used primarily in laboratory-type
investigations and others are used to record behavior in its natural setting
...
Three of the devices used in noncontrived situations are the video camera,
and the Audimeter
...
Eye/pupilometric
cameras, for example, are specially designed cameras that record the pupil size changes and eye
movements to specific materials on a page
...
Subjects are then given an advertisement or other print/text media
...

Pupil size fluctuations are also recorded
...

Interesting or arousing stimuli result in the dilation of the pupil
...

The psycho galvanometer measures the extent of the subject “responds” to the
advertisement
...
This relationship is measured as the
change in electrical resistance in the palms of the subject’s hands
...
In
general, all such devices have the advantage of permitting careful and detailed observations of
behavior that could not be made otherwise
...
In using these devices, however, one should always
keep in mind two important questions:
1
...
Are the subjects behaving as they would in a natural situation?
The answer to the second question can clearly be in the affirmative if the observation is
made outside the laboratory and in the natural situation, such as in ethnographic research
...

Direct Observation
Direct observation of people and how they behave in situations of interest is a commonly
used method of collecting information
...
Supermarkets and
department store managers continually rely on observation of traffic flows and length of waiting
lines to determine the proper location of the various lines of products and the number and
location of salespeople and cash registers
...

115
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Information obtained from direct observation of purchasers can be highly useful in
helping to answer such questions as
Who actually buys the product?
Do they appear to be influenced by an accompanying person?
To what extent do brand choices appear to have been made earlier versus at the point of
purchase?
What proportion of shoppers appears to check prices?
What proportion of shoppers studies the package before purchase?
Unobtrusive Measures

Observation is the method of data collection underlying a set of methods known as
unobtrusive measures
...
Included are all the
types of data collection mentioned in this section
...
Before this change was made, information was needed on the extent to
which consumers actually read information on labels
...
An analysis of these pictures indicated
that far more time was spent in the selection and more careful attention given to the label
than had previously been believed to be the case
...

In addition, other types of unobtrusive measures are traces and archives
...
Garbage data can be used to examine
many aspects of consumption behavior (Reilly, 1984):
1
...

3
...

5
...


Brand and product type switching patterns
Market share estimation
Lifestyle patterns
Ethnic and social group differences
Media usage patterns
Free samples, deal packs, trial sizes, coupons

As an example, consider a “classic” study where alcoholic beverage purchases are
tracked in a “dry” community
...
In this type of unobtrusive trace
research, there are some biases that can arise
...
In general, biases may include the following:
Missing evidence
...

Incomplete evidence
...

116
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Despite these limitations, it appears they are minor relative to the advantages, particularly
when the refuse analysis is part of an accepted project
...
Under the current practices used (random sampling of household
refuse within selected neighborhoods), individuals are not aware that their refuse is being
analyzed
...
Because they are not aware that their consumption is being monitored,
consumers are unlikely to alter their behavior to appear more rational, more socially
acceptable, or more economical
...
There is no selective bias in participation
...

Interviewer effects
...
There is limited
possibility for the recorder to consciously or unconsciously bias the outcome of the
analysis
...
As a result of the unobtrusiveness of the procedure, respondents are
not capable of misrepresenting their behavior, either because they can’t remember
accurately or because they wish to create some type of favorable impression
...
The behavior of the same household can be observed over time
...

Satisfaction
...

Completeness
...
Illegal behaviors, purchases from unscanned stores, outlier
behaviors (such as beer consumption at parties), and socially sensitive aspects of
behavior are all amenable to quantitative analysis using refuse evidence
...
Garbage analysis includes accurate measures of when products were
used, as opposed to measures of when the products were purchased
...

For a more detailed discussion of unobtrusive methods, we refer the reader to the writings
of Webb, Campbell, Schwartz, and Sechrest (1966); Sechrest (1979); and Bouchard (1976)
...

This is the process known as triangulation
...
This division reflects in marketing research
the objectivist-subjectivist debate in the behavioral sciences in general
...
Smith and Gerald S
...

Applicability of Indirect Research Techniques
The basic premises leading to the use of indirect research techniques are as follows:
1
...

2
...

3
...

4
...

How valid are these premises? We have already seen that they are valid for some
problems
...
It is more useful to review
situations in which indirect research information might reasonably be sought from respondents
...

Category one is where the information desired is known to the respondent and he or she
will give it if asked
...
If the reason a consumer does not buy brand X tires is because he believes they do
not wear as well as they should, he will willingly say so given the opportunity
...
Matters that are considered to be private in nature, that are believed
to be prestige- or status-bearing, or that are perceived as presenting a potential respondentinvestigator opinion conflict may not be answered accurately
...
When underlying motivations of this general nature are believed to exist, indirect
techniques are well suited to elicit such information
...
When respondents have reasons they are unaware of, such as the
association of the use of instant coffee with lack of planning and spendthrift purchasing, properly
designed and administered indirect techniques can be highly useful for uncovering such
motivations
...
In some cases motivations of respondents are so deep-seated that neither
direct nor indirect methods of questioning will bring them to the surface
...
Smith and Gerald S
...

In another experiment, orange-scented nylons were placed on a counter in a department store next
to identical, but unscented, hosiery
...
Questioning of the women who bought the scented hose as to why
they preferred the hose they bought resulted in answers such as “of better quality,” “sheerer,” and
the like
...
It remains, of course, for the analyst to decide in which one
or more of these categories the information he or she requires will fall
...

Sample Selection and Sizes in Qualitative Research
The subject of sampling is considered in detail in the next chapter
...

Sample selection in qualitative research studies has tended to be done on nonprobabilistic
(purposive) bases rather than by probabilistic methods
...
Serious sampling errors can result from purposive sampling and the
extent of the sampling error is unknown
...

A second area of controversy over the samples typically taken in qualitative research
studies relates to their size
...
The use of a small sample in a qualitative research study suggests that the population of
psychological attributes and motivations being sampled is sufficiently homogeneous that only a
limited sample is required to provide an adequate representation of the population
...
To assume that the motivations of a very small group of people
adequately represent those of the population at large is to ignore the high degree of variability
that empirical studies have substantiated
...
Unfortunately, to raise the question is to beg it; no definitive answer can be given
...

An observation does need to be made, however, about the differences in judging validity
by the “clients” of basic research versus those of decisional research projects
...
Judgments of a study’s validity are a highly
impersonal process and one that is seldom urgent
...
Smith and Gerald S
...

Since indirect research methods violate each of these requirements to some extent, there
has been reluctance on the part of some basic researchers to give even tentative acceptance to
unvalidated findings of studies that employ indirect methods
...
In the absence of data that can be used for direct validation,
the basic research project is judged tentatively on the basis of method
...
Rather
than wanting to be assured that the best estimate of a parameter or the most definitive test of a
hypothesis has been made, the client needs information that will assist him or her in making the
best decision possible in the circumstances
...
The client works directly with the researcher
and is able to raise any questions he or she has about the project
...
An assessment of validity of
the findings must be made now; for a decision to await the outcome to determine if the findings
are valid would obviate the very purpose for which the research was conducted
...

Indirect techniques serve several useful purposes in marketing research
...

Included in the supplemental information that is of value is that which suggests hypotheses that
can be tested by direct methods
...
In the indirect
types of interviews, we described the more commonly used projective techniques, including the
third person technique (“what does your neighbor think of ___?”), word association, sentence
completion and depth interviews
...

We then considered the means of obtaining information through observation of people
...
Of particular interest is the use of trace analysis (analysis of post-consumption
garbage) to study behavior
...


120
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

REFERENCES
Berstell, G
...
Study what people do, not what they say
...

Berstell, G
...
(2001, February 26)
...

Marketing News, 35, 45
...
, Frankland, J
...
, & Robson, K
...
Focus groups in social research
...

Bouchard, T
...
, Jr
...
Unobtrusive measures: An inventory of uses
...

Ericsson, K
...
, & Simon, H
...
(1984)
...
Cambridge: MIT Press
...
J
...
Some differences between consumer-based and business-to-business groups require
a special touch
...

Fern, E
...
Why do focus groups work? A review and integration of small group process theories
...
441–451
...
L
...
R
...
Focus groups illuminate quantitative research
...

Gibby, K
...
, Lee, F
...
, Merrill, S
...
(1995)
...

(Field study conducted under the supervision of Professor Scott M
...
) Provo, UT: Brigham Young University
...
E
...
S
...
(1988)
...
Englewood Cliffs, NJ:
Prentice Hall
...
(1993, January 4)
...

Marketing News, 27, FG-2
...
(1950, April)
...
Journal of Marketing, 14, 649–656
...
R
...
Focus groups are divided into 4 distinctive categories
...

LaFlamme, T
...
(1988, August 29)
...

Marketing News, 22, 12
...
(1999, January 4)
...
Marketing News, 33, 20
Mariampolski, H
...
Probing correctly uncovers truth behind answers in focus group
...
W
...
Motivation research and marketing management
...

Olson, Jerry C
...
Hinterhuber and K
...
), Kundenorientierte Unternehmensführung, 6th
edition, Gabler Verlag: Wiesbaden, 2008
...
olsonzaltman
...
pdf
Jerry C
...
olsonzaltman
...


121
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Reilly, M
...
(1984, September/October)
...
American Behavioral
Science, 28, 115–128
...
J
...
(1988, February/ March)
...

Journal of Advertising Research, 28, 11–31
...
J
...
, & Westberg, S
...
Advancements in laddering
...
J
...
C
...
, Understanding Consumer Decision Making: A Means-End Approach to Marketing and Advertising Strategy
...

Reynolds, T
...
, & Whitlark, D
...
Applying laddering data to communications strategy and
advertising practice
...

Rydholm, J
...
Preserving the preservationists
...
8–19ff
...
, Ed
...
Unobtrusive measurement today
...
(1988, August 29)
...
Marketing News, 22, 47–
48
...
, Campbell, D
...
, & Sechrest, L
...
Unobtrusive methods: Nonreactive research in the
social sciences
...

Zaltman, G
...
How consumers think
...


122
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 7
SAMPLING PROCEDURES IN RESEARCH
Researchers must answer many questions when developing a project to collect
information about a population of interest
...
This
chapter first considers the selection of sample types and the process of sample planning
...
Finally, related to design, but a unique decision area in its
own right, is the determination of the size of the sample
...

Each involves making inferences about a population on the basis of information from a
sample
...
However, as Exhibit 7
...

Exhibit 7
...
When a
researcher speaks of sampling error and the size of the confidence limits placed on an estimate, that
researcher is speaking of precision and not accuracy
...
The
researcher must constantly ask: “Have we asked the right questions?” and “Have we measured
what we want to measure properly and correctly?”
The difference between precision and accuracy is analogous to the difference between
errors of degree and errors of kind
...
If
a researcher has made the correct measurements, precision will show the degree of error;
but if the right measurements have not been taken, precision is probably meaningless
(Semon, 2000)
...


123
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Consequently, strict attention must be paid to the planning of the sample
...
The
process of selecting a sample follows the well-defined progression of steps shown in Figure 7
...

Figure 7
...
A
population, also known as a universe, is defined as the totality of all units or elements
(individuals, households, organizations, etc
...

While seemingly an easy task, an imprecise research problem definition often leads to an
imprecise population definition
...
For example, a group medical practice that is considering
expanding into sports medicine might acquire information from any or all of the distinct
population groups listed in Table 7
...
The population element is the unit of analysis, and may be
defined as an individual, household, institution, patient visit, and so on
...
1 Possible Population Choices for a Research Study by a Medical Practice

Group Members

Where

When

All patients
Patients who have had orthopedic work
All people
All people who have had orthopedic work

Designated group practice
Designated group practices
Specified geographic area
Specified geographic area

Last 12 months
Last 12 months
Last 12 months
Last 12 months

The second and third columns of Table 7
...
One useful approach is to first define the ideal population to meet study
objectives, and then apply practical constraints to ultimately limit and define the study
population
...
Over defining can limit the extent to which findings can be generalized, and greatly
increase the operational cost and difficulty of finding population elements (Sudman, 1976)
...
Smith and Gerald S
...
The
desirability and advantages of using a sample rather than a census depend on a variety of factors
such as geographic location, the absolute size of the population, and the sample size required for
results sufficiently accurate and precise to achieve the required purposes
...
A
survey based on a sample takes much less time to complete than one based on a census
...

In other situations, a sample is necessary because of the destructive nature of the
measurement, such as in product testing
...
Nonprobability sampling techniques explicitly protect against this
problem
...
Samples (a smaller
number of interviews compared to a census) may result in better interviewing, higher response
rates through more call backs, and better measurement in general
...
In Chapter 2, we emphasized the importance of minimizing total error
...
When the
population is small, the variance in the characteristic being measured is high, the cost of error is
high, or the fixed costs of sampling are high, sampling may not be useful
...
In this
case, it would be necessary to sample a relatively large proportion of the general population to
provide statistically reliable information
...

Sample Design
Sample design is “the theoretical basis and the practical means by which data are collected so that the
characteristics of a population can be inferred with known estimates of error” (Survey Sampling, Inc
...
Sample design
specification, including the method of selecting individual sample members, involves both
theoretical and practical considerations (such as cost, time, labor involved, organization)
...

2
...

4
...

Are rigorous sampling methods chosen? This involves selecting an appropriate
probability or nonprobability sampling method
...
Smith and Gerald S
...
In contrast, probability samples offer the promise of
bias-free selection of sample units and permits the measurement of sampling error
...
In nonprobability sampling one must rely
on the expertise of the person taking the sample has selected respondents in an unbiased manner,
whereas in probability sampling the sampling of respondents is independent of the investigator
...
The
employees’ dogs liked the food and the pet food manufacturer actually introduced the new dog
food product
...
As managers scrambled to find out what went wrong, research showed that employees
were so loyal to the company’s products that their dogs had little variety and would eat anything
for a change
...
In this case, a biased sample was
erroneously assumed to conform to the general population of dogs and dog owners
...
For a given cost, one can usually select a
larger nonprobability sample than probability sample, meaning that the sampling error should be
lower in the nonprobability sample, but that the nonrandom process used for selecting the sample
may have introduced a selection bias
...
The sampling unit may contain one or more elements describing the population
...
In either case, it may be preferable to select a sample
of households as sampling units
...
The sampling frame usually is a physical listing of the population elements
...
For instance, in a consumer survey where personal interviews are
conducted with a mall intercept method is used to obtain data, the sample frame includes all
those people who enter the mall during the study period
...
Smith and Gerald S
...
Such a perfect frame is seldom available
for marketing research purposes
...
2, a sampling frame may be incomplete,
too comprehensive, or a combination of both
...
Any of these situations can lead to coverage error
...
2

Sampling Frame – Population Relationship

Perhaps the most widely used frame in survey research for sampling human populations
is the telephone directory
...
For example, more than one in five American
homes (20
...

Combined with the fact that in many major cities more than 50% of home phones are unlisted,
the potential for frame errors is huge
...
The National Do Not Call Registry provisions do
not cover calls from political organizations, charities, telephone surveyors, or companies with
which a consumer has an existing business relationship, nonetheless, resident push back leads to
further frame error
...
Dillman (2000) suggests asking the following five questions about any potential
sampling list:
Does the list contain everyone in the survey population?
Does the list include names of people who are not in the study population?
How is the list maintained and updated?
Are the same sample units included on the list more than once?
Does the list contain other respondent information that can be used to customize the
survey or direct survey logic?

127
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Whether the researcher uses their own list, or one purchased from a commercial source,
these questions are equally applicable
...
Four traditional
approaches can determine the appropriate sample size for any given research project:
1
...

3
...


Arbitrarily or judgmentally determined
Minimum cell size needed for analysis
Budget-based
Specifying desired precision in advance of sampling and then applying the
appropriate standard error formula to determine the sample size
...

Costs of Sampling
The sample plan must account for the estimated costs of sampling
...
In reality, costs
from all aspects of data collection are usually considered together
...
1 when the sample design and sample size are considered, the estimated costs may be so
great that other sample designs and/or smaller-sized samples may be considered
...
A sample is chose
that is thought to be both representative and to adequately mirror the various patterns and
subclasses of the population (Exhibit 7
...
The sample should be of sufficient size to provide
confidence in the stability of its characteristics
...
2

Samples Are Not Always an Exact Match

Many researchers feel that the best way to assess the validity of a sample is to compare its
demographic profile (i
...
, distributions of the key demographic characteristics) with a national or
otherwise known profile
...
At the same
time, a poor fit does not necessarily mean that the obtained sample is bad
...
However many
studies have shown that large differences in demographic characteristics may to translate into
small differences in the variable of interest, whether it is a behavior or an attitudes, interests, or
opinions (AIO) measure
...
But the differences generally have to be much greater than we would think for a
significant effect
...
From this discussion, one might conclude that an ideal sample is obtained with a
probability process
...
However, recognize that it is more important to
128
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

avoid distorted samples than to be able to measure sampling error
...

This chapter is primarily concerned with probability sampling, but before discussing this
topic it is useful to describe some of the procedures by which nonprobability samples are taken
in marketing research
...
As such, sampling error is generally not able to be measured
...

The Quota Sample
The quota sample is the most commonly employed nonprobability sampling procedure in
marketing research
...
This might be, for example, the proportion of the adult
population who fall into various age-by-gender-by-education groupings
...
Next, if an interviewer has a total number of, say, 600 interviews to
obtain, the age-gender-education proportions in the population are applied to the 600 total
interviews to determine the appropriate quotas
...
The lack of random selection is the major
distinction between quota sampling and stratified random sampling
...
Therefore, many sources of selection bias are potentially present
...
”
Interviewers may go to selected areas where the chances are good that a particular type of
respondent is available
...
” Still other interviewer habits and biases exist that can
influence their selection of respondents within the quota
...
Quota sampling is quite close to
traditional probability sampling when selection is tightly controlled (Sudman, 1976)
...
A
critique of the method is offered by Deming (1960)
...

The Judgment Sample
A somewhat representative sample may be provided through the use of purposive, or
judgment, sampling
...
The intent is to
select respondents representative of the population in such a way that errors of judgment in the
129
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

selection will cancel each other out
...
One weakness of this approach is that without an objective basis for making the
judgments, there is no way of knowing whether the so-called typical cases are, in fact, typical
...
Convenience sampling means that the sampling units are accessible,
convenient and easy to measure, cooperative, or articulate
...
In this case, the researcher needs to select respondents
with care
...
A sample is said to be length
biased if the probability of observing an individual at a particular site is dependent on the
individual’s length of stay at the site
...
Again, the purpose is to
obtain a relatively large number of interviews quickly from a cooperating group of respondents
...
However depending on the purpose of the research, many
potential sources of selection bias may exist in that only certain members may respond, who may
be disproportionately different on one of many demographic, attitudinal or behavioral
dimensions
...
One major advantage of snowball sampling is the ability to estimate various
characteristics that are rare in the total population
...

Example: A study of international tourism, required researchers to interview respondents in the
United Kingdom, France, and Germany who visited the United States in its bicentennial year
...
A referral procedure (up to two
referrals per qualified respondent) was then used to obtain a second group of qualified
respondents
...
)

In other types of snowball sampling, referrals from referrals are obtained, and so on, thus
leading to the “snowballing” effect
...

Major advantages of this type of sampling over conventional methods are the
substantially increased probability of finding the desired characteristic in the population, and
lower sampling variance and costs
...
Smith and Gerald S
...

However, many occasions in marketing research require more specialized sampling procedures
than those of simple random-sampling methods
...
Five major modifications can be made
to the basic selection process, as shown in Table 7
...

These techniques are discussed in turn, following a review of simple random sampling
...
Many excellent statistics and research books
review the mathematical aspects of these sampling techniques
...
It is drawn by a random procedure from a sample frame—a list
containing an exclusive and exhaustive enumeration of all sample elements
...

Simple random samples are not widely used in consumer research, for two reasons
...
Consumer research usually requires people, households, stores, or areas to be the basic
sampling units
...
When persons, households, or stores are to be sampled, some other sample design
must be used
...
In this case purchasing agents, companies, or areas are the usual sampling
units and the population under study is often relatively small
...

Table 7
...
Equal probability for all elements

Unequal probabilities for different elements; ordinarily
compensated with inverse weights

a
...
Element Sampling: single stage, sampling

a
...


Disproportionate allocation designed for
optimum allocation

b
...
Smith and Gerald S
...
One-stage cluster sampling
b
...
Equal clusters
d
...
Unstratified Selection: sampling units
Stratified Sampling: separated selections from partitions,
selected from entire population
or strata, of population
IV
...
One-Phase Sampling: final sample selected Two-Phase (or Double) Sampling: final sample selected
directly from entire population
from first-phase sample, which obtains information for
stratification or estimation
SOURCE: Adapted from Kish, 1965, p
...


The Systematic Sample
Systematic sampling involves only a slight variation from simple random sampling
...
The
permissible samples of size n that are possible to draw each have a known and equal probability
of selection, while the remaining samples of size n have zero probability of being selected
...
If the population contains
N ordered elements and a sample size n is desired, one merely finds the ratio of N/n and rounds
to the nearest integer to obtain the sampling interval
...
A random number is
then selected between 1 and 10, inclusively; suppose the number turns out to be 4
...

Essentially, systematic sampling assumes that population elements are ordered in some
fashion—names in a telephone directory, a card index file, or the like
...
In other instances, the ordering may be related to the characteristic
under study, as when a customer list is arranged in decreasing order of annual purchase volume
...
We say “close” because in systematic sampling, all combinations of
the characteristic do not have the same chance of being included
...

Systematic sampling may increase the sample’s representativeness when items are
ordered with regard to the characteristic of interest
...
On the other hand, the simple random sample
may yield, say, only low-volume customers, and may thus be unrepresentative of the population
being sampled if the characteristic of interest is related to purchase volume
...
For example, a sample
interval of 7 for a systematic sampling of daily retail-store sales figures would reflect the same
day of data for each week and would not reveal the day-of-the-week variations in sales
...
Smith and Gerald S
...
g
...
In such cases,
a separate sample is then taken from each stratum
...
In
practice, however, systematic and other types of random samples are sometimes taken from each
of the strata
...

Stratified samples are generally conducted according to the following procedure:
The entire population is first divided into an exclusive and exhaustive set of strata, using
some external source, such as census data, to form the strata
...

From each separate sample, some statistic (such as a mean) is computed and properly
weighted to form an overall estimated mean for the whole population
...


The two basic varieties of stratified samples are proportionate and disproportionate
...
In disproportionate stratified
sampling, one departs from the preceding proportionality by taking other circumstances, such as
the relative size of stratum variances, into account
...
The researcher may be willing to assume that, although average consumption would vary
markedly by family size, the variances around the means of the strata would be more or less equal
among family sizes
...


More generally, however, both means and variances will differ among strata
...
In this instance, the
number of families included in each stratum would be proportionate to (the product of) the
relative size of the different family-sized strata in the population and the standard deviation of
each family class
...
Formulas for computing sampling errors in stratified samples are briefly
discussed later in this chapter and can be found in standard texts on sampling
...
The greater the within-stratum homogeneity and
among-stratum heterogeneity, the more efficient the stratified sampling is relative to simple
random sampling
...
First, a
sample size needs to be calculated for each stratum or subgroup
...

133
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Nevertheless, this method continues to be widely used due to the segmentation of markets that
companies routinely engage in
...
However,
when larger primary sampling units are desired, cluster sampling may be used
...
The blocks, not the individual families, would be selected at random
...
The main advantage of a cluster sample relative to simple
random sampling is in lower interviewing costs rather than in greater reliability
...
A single-stage area
sample occurs when only one level of sampling takes place (such as a sampling of blocks) before
the basic elements are sampled (the households)
...

Example: Consider the sample design used by the Gallup Organization for taking a nationwide
poll
...
Blocks
or geographic segments are then randomly sampled from each of these locations in a second
stage, followed by a systematic sampling of households within the blocks or segments
...


METHODS FOR DETERMINING SAMPLE SIZE
There are several ways to classify techniques for determining sample size
...
1
Although the discussion will focus on the statistical aspects of setting sample size, it
should be recognized that nonstatistical dimensions can affect the value of a research project
...

Fixed Versus Sequential Sampling
As the name implies, in fixed-size sampling the number of items sampled is decided in
advance
...
In general, all sample observations are taken before the data are analyzed
...
Rather, the
134
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

analyst sets up in advance a decision rule that includes not only the alternative of stopping the
sampling process (and taking appropriate action, based on the sample evidence already in hand)
but also the possibility of collecting more information before making a final decision
...

In general, sequential sampling will lead to smaller sample sizes, on average, than those
associated with fixed-size samples of a given reliability
...
In addition, the problem may be
such that it is less expensive to select and analyze a sample of many items at one time than to
draw items one at a time (or in small groups) and analyze each item before selecting the next
...
Also, we would expect that the less
dispersed the population’s characteristics are, the closer our sample estimates should be to the
true parameter
...
These inferences should be more reliable when the sample is larger and
when there is less variability in the population variables measured
...
3
...

The “margin of error” or “confidence interval” is the plus-or-minus figure that represents the
accuracy of the reported
...
”
Eighty-two percent of Canadians expect to buy gifts for their mom, compared to 20 percent for
their wife and 15 percent for their mother-in-law
...
The national findings are
accurate, plus or minus 2
...


In this example, if 82% of your sample indicates they will "buy a gift for mom" and you
use a confidence interval of 2
...
25% (82%-2
...
75% (82%+2
...

The “confidence level” tell you how confident you are of this result
...
The 95% confidence level means that if 20 different samples were drawn, 19 times out of
20, the results would fall in this - + confidence interval
...
The 95%
confidence level is the most commonly used
...
25% and 84
...

Wider confidence intervals increase the certainty that the true answer is within the range
specified
...
When the costs incurred from making an error are
135
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

extremely high (you are betting your company, or a multi-million dollar decision is being made)
the confidence interval should be kept small
...

Exhibit 7
...
election polls is that they are straightforward
...
No
spin
...
The difficulty of
...
How else could you explain that a poll done by one candidate shows
him in the lead and that a poll done by his opponent shows her in the lead? Statisticians say
there are ways to twist questions or interpret answers to give one candidate an advantage over
another
...
He
questioned whether the methodology was described in enough detail, whether the sample size
was adequate
...
The story did not make clear who was polled
...
It should have indicated that the people polled were
registered and likely to vote in the November elections
...
He said the
sample size of 625 likely voters was insufficient for a state with nearly 4 million residents and
suggested at least 800 should have been polled
...
It produces, as the story stated, a margin of
error of plus-or-minus 4 percent
...
5 - more accurate, but not so much more accurate to justify
the additional cost
...
They believe that, the larger
the pool, the larger the sample size needs to be
...
You can take a drop of blood
from a 400-pound person and it will contain the same data you would get if you took it from a
100-pound person," he said
...
"If 'minorities' constituted 27 percent of the total sample,
then only 169 were sampled
...
" Coker said
the reader is right and wrong
...
In this case, it moves from plus-or-minus 4
to plus-or-minus 6 to 8
...
That poll indicated incumbent Gov
...
(Our) poll indicated challenger Mark Sanford was ahead 45 to 41 percent
...


136
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Controlling the Size of the Confidence Interval
Sampling theory teaches us that the accuracy of a sample estimate is dependent on such
factors as the dispersion and skewness of the population’s responses, the sample size, and the
size of the population
...
Note that "non-sampling" errors, such as bad question design or selection of a
"bad" sample frame are not controlled by sample size
...
Larger samples tighten the size of the confidence interval, making your
estimate much more accurate
...
3
...
38 to +- 3
...

Dispersion
The accuracy of an estimate also depends on the dispersion and skewness of the
population on the question being asked
...
Bush
than would a sample of democrats
...
Accuracy of
the sample estimate increases as the dispersion in the population decreases
...

When using a proportion to compute sample size or estimate confidence intervals, it is
easy to cover all eventualities by assuming maximum variability (50-50 proportions)
...

Population Size
The size of the population also influences the size of the confidence interval, but not as
much as you might expect
...
08%
...
1%
...

Non-sampling errors cannot be compensated for by increased sample size
...
Non-sampling errors come
from samples that are not truly random, bad scales, misleading questions, incomplete surveys,
etc
...
Smith and Gerald S
...
3

Proportion Based Confidence Intervals Computed at the 95% Confidence Level
Sample
Size

Variability Proportions
50/50% 40/60% 30/70% 20/80% 90/10%

95/5%

25

20

19
...
3

16

12

8
...
2

13
...
4

8
...
2

75

11
...
3

10
...
2

6
...
8

9
...
4

150

8
...
5

6
...
9

3
...
1

7

6
...
7

4
...
1

250

6
...
2

5
...
8

2
...
8

5
...
3

4
...
5

2
...
9

4
...
2

500

4
...
4

*4
...
6

2
...
1

4

3
...
3

2
...
8

800

3
...
4

3
...
8

2
...
5

1000

3
...
0

2
...
5

1
...
4

1500

2
...
5

2
...
0

1
...
1

2000

2
...
2

2
...
6

1
...
96

2500

2

1
...
8

1
...
2

0
...
4

1
...
3

1
...
83

0
...
1
percentage points at the 95% confidence level
...


Sampling Basics: Sampling Distributions and Standard Errors
The reader will recall from elementary statistics the concept of a sampling distribution
...
g
...
The standard error of the statistic is the standard deviation of the specified sampling
distribution
...
Smith and Gerald S
...
The arithmetic mean of the sampling distribution of the mean ( X ) or of the proportion (p)
for any given size sample equals the corresponding parameter values

and

, respectively
...
The sampling distribution of the means of random samples will tend toward the normal
distribution as sample size n increases, regardless of the original form of the population being
sampled
...
For large samples (e
...
, n 100 and for fairly close to 0
...

4
...
At a
10% sampling fraction, taking it into account will reduce the random sampling error by 5%
...
5%, 1
...
5%, respectively
...
Probabilities of normally distributed variables depend on the distance (expressed in multiples
of the standard deviation) of the value of the variable from the distribution’s mean
...
In symbols this is as
follows:
Zi

Xi

Table A
...
Note further that the original variate may be a sample mean, X
...
e
...
We
can then define Z as some number of standard errors away from the mean of the sampling
distribution
Z

X
X

where

X

denotes the standard error of the mean
...
)
6
...
Smith and Gerald S
...
If the population standard deviation
is not known, which is often the case, we can estimate
it from the sample observations by use of the following formula:
n

s

(X1

X )2

i 1

n 1

We can consider s to be an estimator of the population standard deviation,
...
g
...
2 in Appendix A is appropriate for
finding probability points
...
1 is a good approximation of the t distribution
...
X
n
is estimated from s, as defined in the preceding
where est
...

8
...
P
n
as an estimated standard error of the proportion
...
However, if n exceeds about 100 (which is typical of the samples obtained in
marketing research), this adjustment makes little difference in the results
...

First, the analyst can simply select a size either arbitrarily or on the basis of some
judgmentally based criterion
...

Second, analysis considerations may be involved and the sample size is determined from
the minimum cell size needed
...

Third, the budget may determine the sample size
...

It may appear that these methods are for nonprobability samples
...
Smith and Gerald S
...
For probability samples, the precision must be determined after the fact
...
This approach using traditional inference (Neyman-Pearson), relies on either of two
major classes of procedures available for estimating sample sizes
...
This can be called the confidence-interval approach
...
We discuss each of these approaches in turn
...
First, as with the other approaches,
the analyst must still calculate the standard error after data collection in order to know what it is
for the actual sample that provided data
...
Depending on the data collection method
used, the original survey delivery may have to be much larger
...
A mail survey is used for data collection and past experience
has shown that the response rate would be around 25%
...

The Confidence Interval Approach
It is not unusual to construct a confidence interval around some sample-based mean or
proportion
...
For example, suppose a
researcher sampled 100 student consumers and noted that their average per capita consumption
of specialty fruit/energy drinks was 2
...
Past studies indicate that the population
standard deviation
can be assumed to be 0
...

With this information, we can find a range around the sample mean level of 2
...

For example, suppose that we wished to set up a 95% confidence interval around the
sample mean of 2
...
We would proceed by first computing the standard error of the mean:
0
...
03

From Table A
...
96 Z variates (2
...

With this information we can then set up the 95% confidence interval as
X

1
...
6 1
...
03)

and we note that the 95% confidence interval ranges from 2
...
66 pints
...
54 and 2
...

This basic idea can be adapted for finding the appropriate sample size that will lead to a
141
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

certain desired confidence interval
...
The researcher knows that it is possible to take a
random sample of respondents in the area and compute a sample mean
...
The researcher
wants to find a sample size that will permit this kind of statement
...
5 gallon of energy drinks per capita and
the level of confidence is 95%
...
Specify the amount of error (E) that can be allowed
...
X E , therefore,
will lie with some prespecified level of
defines the interval within which
confidence
...
5 gallon per year
...
Specify the desired level of confidence
...
Determine the number of standard errors (Z) associated with the confidence level
...
For a
95% confidence level, reference to Table A
...
025 probability that the population mean will fall outside one end of the interval is
Z = 1
...
Since we can allow a total probability of 0
...
96 is the correct value for a 95%
confidence level
...
Estimate the standard deviation of the population
...
Suppose that the standard deviation of the area’s population for specialty
fruit/energy drink consumption is assumed to be 4
...

5
...
One
standard error of the mean is to be set equal to the allowable error (E = 0
...
96
...
5
0
...
96
This will assure us that the interval to be computed around the to-be-found sample mean will
have a 95% preassigned chance of being ±0
...

6
...
255

4
...
Smith and Gerald S
...
In general, we can find n directly from the following formula:
2

Z 2 16(1
...
5) 2
If the resulting sample size represents a significant proportion of the population, say 10% or
more, the finite population multiplier is required and the sample size must be recalculated
using the following formula, where N is the size of the population:
n

2

N(
NE 2

n

Z 2)
2 2
Z

The Case of the Sample Proportion
Suppose that, in addition to estimating the mean number of gallons of specialty
fruit/energy drinks consumed per capita per year, the researcher is also concerned with
estimating the proportion of respondents using one or more specialty fruit/energy drinks in the
past year
...
In this case the following checklist would be
used:
1
...
Suppose that the desired reliability is such
that an allowable interval of p
0
...
05, or 5
percentage points
...
Specify the desired level of confidence
...

3
...
This will be
the same as for the preceding estimation; Z = 1
...

4
...
The population proportion can again be estimated
by judgment, by reference to other studies, or by the results of a pilot sample
...
4 in this case; that is, the researcher assumes that 40% of the population
used one or more specialty fruit/energy drinks last year
...
Calculate the sample size using the formula for the standard error of the proportion
...
05) divided by
the appropriate Z value of 1
...

P

E
Z

0
...
96

0
...
Neglecting the finite multiplier, we then solve for n in the following formula:
P

(1
n

E
Z

)

0
...
4(0
...
In general, we can find n directly from the formula
n

)Z 2

(1
E2

0
...
4)(1
...
05) 2

369

Once again, if the resulting n is 10% or more of the population size, the finite population
multiplier is required and the sample size can be computed from the following:

143
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

n

N (1
)Z 2
NE 2
(1
)Z 2

Determining Sample Size When More Than One Interval Estimate Is to Be Made
from the Same Sample
The usual case when collecting sample data for estimation of various parameters is that
more than one estimate is to be made
...
Since only one sample is to be chosen, what size should it be?
A strict adherence to the allowable error and the confidence levels specified in the
calculation of the sample sizes for the individual estimation problems leaves no choice but to
take the largest sample size calculated
...
In the specialty fruit/energy drink consumption problem, for example, the sample size
would be 369 (the sample size calculated for estimating the proportion of users) rather than 246
(the sample size calculated for estimating the mean amount used)
...
In order to determine the size of the original
sample the researcher must estimate the rate of response expected
...

Devices for Calculating Sample Size
In practice, manual devices (paper nomographs) and online sample size calculators can
be used to easily compute a sample size helpful in rough-guide situations where the researcher is
not sure of either allowable error levels or population standard deviations
...
byu
...
html
...
In this case the procedures
are more are more elaborate
...
These risks are, in turn, based on H0: the null hypothesis, and H1: the alternate
hypothesis
...

This is a type I error
...
This is a type II error
...
4
...
We first consider the case for
means and then the case for proportions
...
The relationship between these
two errors is an inverse
...


144
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Table 7
...
b
...
4
Action
Accept Ho
Reject Ho

Types of Error in Making a Wrong Decision
Ho is true
No error
Type I error ( )

Ho is false
No error

When the hypothesis is one of difference, a type II error occurs when what is really
chance variation is accepted as a real difference
...
Sample size is affected by the
effect size—the needed or expected difference
...
Very small effect sizes (e
...
, two percentage
points) require such large sample sizes as to be impractical; surveys cannot really reliably
measure such small differences or changes
...
If the desired power is raised to 95%, the sample size needed would be about 540 (Semon,
1994)
...

145
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The Case Involving Means
As an illustrative example, let us assume that a store test of a new bleaching agent is to be
conducted
...
On the other hand, a mean sales level
of 10 cases per week would justify marketing the new product nationally
...
Specifically, we go through the following checklist:
1
...
(By convention, the null
hypothesis is the one that would result in no change being made, if accepted
...


2
...
The type II error is
allowable probabilities of making those two types of errors, respectively
...
3, where we assume that in the bleach-introduction
Figure 7
...
Determine the number of standard errors associated with each of the error
-tailed test the Z values for the 0
...
01 risks,
respectively, are found from Table A
...
64 and Z = 2
...

These are shown in figure 7
...
Note that in the figure we affix a minus sign to the
value of Z since the critical values lies to the left of 1 = 10
...
Smith and Gerald S
...
Estimate the population standard deviation
...

5
...
Because two
sampling distributions are involved, a simultaneous solution of two equations is
required to determine the sample size and critical value that will satisfy both
equations
...
Setting the right-hand side of these two equations equal and solving for n gives
(Z
Z )2 2
n
2
( 1
0)
In the bleach problem the desired sample size is
n

(1
...
33) 2 5 2
(10 7) 2

44 stores (rounded)

Having solved for n, the sample size, we can then go on to solve for the critical value for the
mean number of cases by means of the following substitution:
critical value
Z
1
n
5 = 8
...
33)
44

Alternatively, we could find the critical value from the first of the two equations:
critical value = 7 + (1
...
24

The decision rule then becomes the following: Take a sample of 44 stores for the
controlled store test
...
24 cases, do not introduce the product
...
24 cases, introduce the product
...
Specify the values of the null (H0) and the alternate (H1) hypotheses to be tested in terms of
population proportions 0 and 1 , respectively
...
Specify the allowabl
3
...

4
...
Smith and Gerald S
...
To illustrate its application, suppose that a researcher is
interested in the true proportion of residents in a large city who would be willing to pay over
$400 for a portable refrigerator-bar combination if it were commercialized
...
If the proportion is only 60%, the researcher would not recommend
commercialization
...
6

H1:

1=

0
...
05 if the true proportion is equal to 0
...
Moreover, the researcher is willing to assume a beta
risk of 0
...
7
...
05 = 1
...
1 = 1
...
6, and

1=

1

(1

1

)

2

0

0
...
The solution follows:
2

1
...
6(0
...
28 0
...
3)
0
...
6

193 (rounded)

Accordingly, in this example the sample size to take is 193
...
7 (1
...
7(0
...
658
193

In this case the decision rule is the following: Take a sample of 193 residents
...
658, do not commercialize the
refrigerator-bar combination
...
658, commercialize the product
...
How are the sizes for other types of randomsample designs—systematic, stratified, cluster, area, and multistage—determined?
148
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The answer to this question is that the same general procedures are used to determine the
overall sample size, but the formulas for the standard errors differ
...
This is because the standard error for these designs is partially a
function of the standard deviation (or proportion) of each stratum or cluster included in the
design
...
And once the overall sample size is determined; it must be apportioned among the strata
and clusters, which also adds to the complexity
...
In general, as compared
with the size of simple random samples, systematic samples may be the same (for purposes of
calculating the standard error, the assumption is typically made that the systematic sample is a
simple random sample)
...
If used properly,
stratification usually results in a smaller sampling error than is given by a comparable-size
simple random sample
...
67):
For the same level of precision, one would need a smaller sample size in total, and
this leads to a lower cost
For the same total sample size, one would gain a greater precision for any
estimates that are to be made

EVALUATION OF THE TRADITIONAL APPROACH
If one were to devise the ideal method of determining sample size, at a minimum one
would want it to meet the criteria of being:
1
...

3
...

The traditional approach is logically incomplete because sample size is specified as being
a function only of the conditional probabilities of making errors
...
More advanced texts, however, do consider traditional sample-size determination by
means of formulas that include the costs of sampling (see Cochran, 1963, Chapter 4)
...
However, the only way that accommodation can be made is through
adjustment of either the specified confidence level or the assigned alpha and beta risks
...
Smith and Gerald S
...
We discussed the determination of sample size from the standpoint of
traditional inferential methods and all aspects of sample size determination were covered
...
We concluded the chapter with
an evaluation of this traditional approach to determining sample size
...
G
...
Sampling techniques (2nd ed
...

Deming, W
...
(1960)
...
New York: Wiley
...
A
...
Mail and internet surveys: The tailored design method (2nd ed
...

McDowell, Elisa (2002), "Inside the paper's election polls", The Charleston Post and Courier
Accessed August 21, 2009 http://www
...
com/coms2/summary_02862181489_ITM
Kish, L
...
Survey sampling
...

National Center for Health Statistics (2009), Data: Cell Phone-Only Households State-by-State,
Associated Press, as reported on Fox News, March 11, 2009
http://www
...
com/story/0,2933,508926,00
...
cdc
...
htm
http://marketing
...
edu/samplesizecalculator
...
qualtrics
...
(2000, April)
...
Quirk’s Marketing Research Review, pp
...

Semon, T
...
(1990, January 8)
...

Marketing News, 24, 31
...
T
...
Save a few bucks on sample size, risk millions in opportunity
loss
...

Semon, T
...
(2000, June 9)
...
Marketing
News, 34, 11
...
(1976)
...
New York:Academic Press
...
Smith and Gerald S
...
Quota versus probability sampling
...

Tatham, E
...
A report on sample size selection
...

Unlisted phone numbers on the rise (1989)
http://findarticles
...
, Burton, S
...
(1983)
...
In W
...
Darden et al
...
), Research methods and causal models in marketing
...


151
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

This page is intentionally blank

152
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

CHAPTER 8
EXPERIMENTATION
Experimentation is widely used in marketing research
...
For example, as shown in Exhibit 8
...

This chapter discusses the objectives of experimentation illustrates techniques for designing and
analyzing marketing experiments including:
the nature of experimentation
ingredients of a marketing experiment
sources of invalidity
models of experimental design
panels and experimental design
difficulties in field experiments in marketing
This chapter further discusses experimental designs and logic for advanced surveys using the
capabilities of Qualtrics
...

The Nature of Experimentation
Two general types of experimental designs exist—natural and controlled
...
“Nature”
produces the changes
...
manipulation of at least one assumed causal variable
2
...

True experiments have both types of intervention, while quasi-experiments manipulate
the variables but do not randomly assign the subjects
...
e
...
Quasi-experiments, on the other hand, have treatment, outcome measures, and
experimental units but do not randomly assign subjects to treatments
...
Smith and Gerald S
...

Exhibit 8-1 Test Audiences have Profound Effect on Movies
What if E
...
hadn't made it home, or Richard Gere had failed to come back for his pretty woman,
Julia Roberts? Believe it or not, in the original versions of these films, E
...
died on American soil and it
was Roberts who rejected Gere at the end of "Pretty Woman
...
In what may be Hollywood's last and most closely guarded
secret, the test audience is having a profound effect on the movies you watch
...
But test audiences wanted more
...
That's what can happen if test
audiences love you
...

Director Ron Howard and his partner, producer Brian Grazer, are responsible for hits like Far and
Away, Ransom, Apollo 13, and A Beautiful Mind
...
"
Grazer and Howard
were dealt one surprise when they tested the 1989 film Parenthood
...
"We took out a lot of the vulgarity
...
It didn't offend anybody
...
In
the discussion of this chapter we shall use the term “experimentation” to describe an experiment
may be conducted for the purpose of identifying relevant variables as well as the functional form
of the model that links these variables with the criterion variable under study
...

Experimentation permits the establishment of causal relationships
...

An Industry Example
A national producer of packaged candies was interested in children’s preference for
various formulations of one of its well-known candy bars
...
Paired
154
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

combinations (i
...
, two combinations at a time) comparisons involving the 18 combinations were
made up and evaluated by various school children between 8 and 12 years of age
...
In addition, while children preferred more peanuts to fewer peanuts, the intermediate
level of caramel was the most preferred
...

Many other experiments have been carried out involving taste testing, package design,
advertising type and quality, price sensitivity, and other marketing variables
...
Our concern in this chapter is primarily with defining variables,
designing the experimental procedure, and conducting the experiment
...

Figure 8-1
...
First, there is the “treatment” variable
whose effect upon some other variable the experiment is designed to measure
...
It is often referred to as the
independent variable
...

When this is the case, the researcher may be interested in observing the effects of combinations
of treatment variables, in addition to the effect of each one individually
...
Interaction refers to the situation like the candy bar example cited above,
where the response to changes in the levels of one treatment variable (type of chocolate) is
dependent on the level of some other treatment variable(s) (amount of caramel) in the
experiment
...

In the preceding candy bar example, the dependent variable was product preference
...
Smith and Gerald S
...
e
...
These are known as
extraneous variables, and unless controlled adequately they are the source of errors in an
experiment
...

Test Objects
The terms test units, test objects, and subjects are used interchangeably in discussing
experimentation
...
In marketing research the experimenter has a choice of three possible
universes of test units—people, stores, and market areas
...
13-15)
...
For example, if a researcher is interested in the effect of shelf height on the sales of a
packaged consumer product, it is to be expected that stores will vary in their amount of shopping
traffic, placement of shelving units, and the like
...
If so, he or she may use a technique called covariance analysis, in which
responses to the controlled variables (shelf height) are adjusted for inherent differences in the test
objects (stores) through measurement of these characteristics before (or during) the experiment
...

Our concern at this point is with the operational problems of measurement
...

In a marketing experiment, it is the outcome or dependent variable that is measured
...
Physiological measures include those obtained from devices that measure such things
as eye movements or electrical conductivity of the skin (psychogalvanometer) during advertising
readership
...
Psychological measures include
verbal measures include spoken and written responses, including responses provided
interactively with a personal computer
...
Such measures are obtained
by self-report, observation, or electronic means
...

There are at least three ways in which variation in the independent variable can be achieved
...
For example, a company interested in the effect of
image on sales could conduct an experiment by running a series of advertisements, each of which
was designed to convey a different image of its product
...

Second, there is the presence versus absence technique
...

Finally, the amount of a variable can be manipulated; different amounts are administered to
different groups
...
”
156
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

As with any approach to marketing research, all phases of an experiment should be
carefully planned in advance
...
This includes the setting of the experiment, physical arrangements, apparatus that
will be used, data collection forms, instructions, recording of the dependent variable, and so
forth
...
Those factors that probably would not influence the
dependent variable can be ignored
...

In any experiment, control over all possible variables affecting the response is rarely
possible
...
But compared with the laboratory situation, the researcher who is working in
the marketplace has a really difficult control job to do
...

There are five general techniques for controlling extraneous variables
...
Elimination of extraneous variables, including controlling the situation in which an
experiment is being conducted so as to keep out extraneous forces
...

2
...
Control of the treatment variable helps
separate out the effects attributable to irrelevancies that are correlated with a treatment
...
Matching (sometimes called balancing)
...
In equating subjects, each group is
controlled for all variables except the treatment variable
...
Matching by holding variables constant involves
creating constancy for all groups
...
If time
of day is an important extraneous variable that could affect one’s “taste buds,” then the
experiment should be conducted at approximately the same time on successive days
...
Similarly, using
the same facilitator in all versions of the study may control experimenter effects
...
Counterbalancing is a technique used to control confounding, or the tangling effects of
two or more levels of a treatment variable (for two or more treatment variables), used in
157
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

experimentation
...
Suppose that a
marketing researcher is interested in conducting a series of taste-testing experiments for a
new soft drink
...
A control procedure called counterbalancing would have each subject
taste each of two drinks on the assumption that ratings will be expressed in terms of
differences in sweetness over each subject
...
” To reduce carryover tendencies, the subject would be
asked to take a sip of water between testing trials
...
Randomization provides assurance that known and unknown extraneous factors will not
cause systematic bias
...
Consequently, it is assumed that the effects of the extraneous
variables will affect all groups in an experiment to the same extent
...
e, into random error
(Brown and Melamed, 1990, p
...

The above discussed techniques for control of extraneous variables are a key part of all
experiments, however the fact remains that confounding effects can never be entirely eliminated
...
Each extraneous force potentially has a bearing on the
validity of an experiment and, consequently may threaten the validity of the results
...
Four distinct types of validity have
been identified (Cook and Campbell, 1990, chap
...
Statistical conclusion validity involves the specific question as to
whether the presumed independent variable, X, and the presumed dependent variable, Y, are
indeed related (Rosenthal and Rosnow, 1991, chap
...
After it has been determined that the
variables covary, the question arises as to whether they are causally related
...
A given experiment is internally valid when the observed effect is due
solely to the experimental treatments and not due to some extraneous variables
...
The third type of
validity, construct validity, is essentially a measurement issue
...
Smith and Gerald S
...
Because construct validity is concerned with
generalization, it is a special aspect of external validity
...
That is,
external validity is concerned with how good an experiment is: to what degree can conclusions
of an experiment be applied to and across populations of persons, settings, times, and so on
...
That is,
ways of increasing one kind may decrease another kind
...
For applied marketing research, it has been observed that the priority
ordering is internal, external, construct of the effect, statistical conclusion, and construct of the
cause (Cook and Campbell, 1990)
...
Construct validity is discussed in Chapter 9, and factors
affecting statistical conclusion validity are covered throughout the later sections of this book
...
The kind of evidence that is
required to support the inference that independent variables other than the one(s) used in an
experiment could have caused the observed effect(s) varies, depending on the independent
variable(s) being investigated
...
The following factors affect internal validity:
1
...

3
...

5
...

7
...


History
...

Maturation
...

Testing
...

Instrumentation
...

Selection
...

Mortality
...

Statistical Regression
...

Interactions with Selection
...


Example: To illustrate each of these factors, suppose that a controlled experiment is
setup for salesperson retraining
...
The brand manager of a particular brand of detergent wants to determine if
retraining is a producer (i
...
, a cause) of sales performance
...

159
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

History deals with events outside the design that affect the dependent variable
...
For salespersonretraining, the level of competitive promotion and advertising, the overall level of demand, or
any one of many other producers may have changed substantially in some territories
...

Maturation is concerned with the changes that occur with the passage of time in the people
involved in the design
...
Similarly, some biological and psychological changes within salespeople and
customers over time can affect the performance of either
...
Familiarity with a measurement (i
...
, a test) can sometimes enhance performance
because items and error responses are more likely to be remembered at later measurement
sessions
...
There is the
possibility that some responses obtained after the retraining program were due to the salespeople
remembering responses given on the first measurement
...
This can occur in many ways, including changes in cost-value (a
drop in home prices) over time, the learning process on the part of the investigators, a change in
investigators, or simply boredom or fatigue may affect the measurements and thus the
interpretation of results
...
In this case, there exists the possibility of a confounding of the instrument
with the experimental treatment because the subjects’ understanding or awareness of the variable
being measured may have changed
...
If the selection procedure is randomized, the effect will be a
measurable random variation
...
Sizable systematic
errors may well result
...

Statistical regression dictates that under conditions where measures are unreliable, high
pretest scorers will score relatively lower at the posttest and low pretest scorers will score higher
...

Some of the foregoing error sources affecting internal validity can interact with selection
to produce forces that might appear to be treatment effects
...
Selection-history can occur when the
experimental groups come from different settings
...
In this situation, each group may have a unique local history that might affect
outcome variables
...
Smith and Gerald S
...

The following are sources of error in external validity:
1
...

3
...


Reactive Effect of Testing (Interaction)
...

Reactive Effects of Experimental Situation
...

Interaction of History and Treatment
...

Interaction of Selection and Treatment
...


A frequent problem in research design is that a “before” measurement is desired, but it is
recognized that making such a measurement may alert the subjects that they are participating in a
study
...
Although this potential source of error may seem the same as that from
testing, it is distinctly different that it is an interaction of testing and treatment
...

Experimental situations refers to “all aspects of the experiment which cause the subject to
perceive, interpret, and act upon what he believes is expected or desired of him by the
experimenter (Sawyer, 1975, p
...
This includes the setting within which the experiment is
conducted, the arrangements made for the experiment such as the apparatus used, and the
presence and behavior of an experimenter
...
7)
...
Ideally, the investigator would
seek subjects who are faithful or naïve (i
...
, do not know the hypothesis)
...
Research design, measurement of dependent
variables, and use of procedures (e
...
, deception and natural environments) will reduce such
demand characteristics (Sawyer, 1975, pp
...
In addition, manipulation and confounding
checks can be used to investigate the plausibility of demand characteristics (Perdue and
Summers, 1986, p
...

History-treatment interaction can affect external validity when the dependent variable
measurement and the treatment variable measurement are not time generalized
...
Moreover, with some independent
variables their effects may be long-run in nature, but typically dependent variable measurement
occurs at one point in time
...
Smith and Gerald S
...
Experiments, including laboratory and Internet based
experiments, rely heavily upon volunteers, paid or otherwise
...
88-90)
...
Experimental designs can be categorized into two broad groups—
classical and statistical
...

Classical Designs
The major types of classical designs were mentioned earlier in this chapter
...
Quasiexperimental designs involve control but lack random assignment of subjects, as required for true
experiments
...

The following notational system will be used in the discussion of classical experimental
designs:
X represents the exposure of test groups to an experimental treatment of a producer or
event whose effect is to be observed and/or measured
...

R indicates that individuals have been assigned at random to differing treatments
...
When O’s and X’s are
found in a given row, they are to be interpreted as having occurred in sequence and to
the same specific individual or group
...

Three classes of designs seem to fit for pre- and quasi-experiments:
Time-series and trend designs
Cross-sectional designs
Combinations of the two previous classes
Time-Series and Trend Designs

Time-series and trend designs are similar in concept, yet their differences in
implementation and analytic procedures warrant a brief discussion
...
The
common method of gathering primary data of this kind is to collect current data at successive
intervals through the use of a continuous panel
...
Smith and Gerald S
...
If the latter technique issued,
respondent recall must be relied on to reconstruct quasi-historical data
...

Trend data differ from time-series data in that they are obtained from statistically
matched samples drawn from the same population over time
...

Both time-series and trend data are used to investigate the existence and nature of causal
relationships based on associative-variation and sequence-of-events types of evidence
...

Time-series and trend designs involve at least one treatment and a subsequent
measurement and can involve a large number of measurements with several interspersed
treatments over a long period of time
...

After-Only without Control Group

This design is often termed a “tryout” or “one-shot case study
...

Symbolically, it may be diagrammed as follows:
X
O
(1)
The many weaknesses of this design may be illustrated by applying it to the salespersonretraining problem
...
A group of salespeople are selected by a nonrandom method and
retrained (X); a measurement (O) is made after the retraining
...
The effects of history,
maturation, and selection are all potentially substantial and non-measurable
...

Before-After without Control Group

This design is the same as (1) with the addition of a “before” measurement
...

Although design (2) is relatively weak, it is frequently used
...
In terms of the
salesperson-retraining illustration, a measurement (O1) of sales volume of the retrained
salesperson is made for the quarter of the preceding year
...
It may be possible to find a comparable, if not equivalent, group to serve as a
control against which to compare the results of the group that underwent the treatment involved
...
Smith and Gerald S
...
Note that the individuals
constituting the groups were not selected at random
...

This design can easily be adapted to the sales-retraining evaluation problem
...
Either a
comparable group or all the rest of the salespeople could be selected as the control group
...

Cross-Sectional Designs

Cross-sectional designs involve measuring the dependent variable of interest for several
groups at the same time, the groups having been exposed to different levels of treatments of the
treatment variable whose effect is being studied
...


...
This design
can be used when direct manipulation of the producer involved is not possible or practical
...
The effect of the
different levels of treatment is measured by determining the degree of association between
producer and product
...

A variation of this design is the static-group comparison
...
There could be a
sizable differential effect of extraneous variables between the groups being measured
...
The multiple-time-series
design (3) could be considered a combination of the two types, as it involves measurements of a
product for different groups at the same time as well as for the same group over time
...
One commonly
used design is the ex post facto test-control group
...
This design is illustrated as follows:
164
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

O1
X
O3
(5)
O2
O4
This design is widely used in connection with testing the sales effectiveness of price
changes, “deals,” and advertising
...
After a given advertisement is run
(X), panel members may be questioned to determine whether or not they saw it
...
Those who
did not see the advertisement become a part of the control group
...

True Experimental Designs

As previously mentioned, for true experiments two kinds of investigator intervention are
required:
Manipulation of at least one assumed causal variable
Random assignment of subjects to experimental and control groups
Through use of a random selection procedure, systematic errors due to selection are eliminated
and the effects of the many extraneous variables tend to be equalized between the experimental
and the control groups as the size of these groups increases
...

Three single-variable experimental designs are described below
...
It requires only
one treatment and an “after” measurement of both the experimental and the control group
...
It is illustrated as follows:
R
X
O1
(6)
R
O2
The absence of a "before" measurement (or "pre-test") is a feature that concerns many
researchers about this design
...

This design is of major interest, therefore, when “before” measurements are impractical or
impossible to obtain and/or when the testing and instrument effects are likely to be serious
...

Random-sampling procedures are used to select an experimental and a control group
...
“After”
measurements of sales to each group are made and the differential is determined (O1 – O2)
...
Smith and Gerald S
...
Most of the sources of
systematic error are controlled in this design
...
If the results of each of these evaluations are consistent, the strength
of our inferences about the effect of the experimental treatment is substantially increased
...
“Before”
measurements can be made on the test and control panels, an experimental advertising treatment
introduced on the test cable, and “after” measurements made for both panels
...
In this section, we have discussed the more common
designs
...

More detailed discussions are found in Cook and Campbell (1990), Campbell and Stanley
(1966), and Shadish, Cook and Campbell (2002)
...
In addition, such designs can examine the effects of more than one
independent variable
...
These are
discussed in more depth in Brown and Melamed (1990) and other specialized experimental
design texts such as Field and Hole (2003) and Shadish, Cook and Campbell (2002)
...

Completely Randomized Design

The completely randomized design is the simplest type of statistical design
...
Any number of
treatments can be assigned by a random process to any number of test units
...
The marketer has been able to obtain an email list of
customers and has corporate approval to run an experiment involving the three ads
...
This could be described
as classification by a single factor
...
Moreover, there is no absolute guarantee that
randomization will keep extraneous influences in check
...
Smith and Gerald S
...
Or, suppose that the researcher would like to generalize the
results of the experiment to other sizes of ads or products
...
That is, not only may single-factor
manipulation be difficult to do in practice, but it may be inefficient as well
...

Sources of Invalidity and Selected Experimental Designs

-

-

-

+

+

+

+

+

Multiple Time-Series
O O O X O O O
O O O
O O O
Static-Group Comparison
X
O
O
Nonequivalent Control Group Design
O
X
O
O
O
Posttest-Only control Group Design
R
X
O
R
O
Separate-Sample Pretest-Posttest Design
R
O
(X)
R
X
O
Pretest-Posttest Control Group Design
R
O
X
O
R
O
O

Reactive Arrangements

-

Interaction of
Testing and X

One-Group Pretest-Posttest Design
O
X
O
Time Series
O O O X O O O

Interaction of
Selection and Others

-

-

-

?

+

+

-

-

-

?

?

+

+

+

+

-

?

?

+

+

+

+

+

+

-

-

?

?

+

+

+

-

-

-

+

+

+

+

?

+

+

-

-

?

?

+

+

+

+

+

+

+

+

+

?

?

-

-

+

?

+

+

-

-

+

+

+

+

+

+

+

+

+

+

+

-

?

?

Testing

Mortality

-

External

Selection

Maturation

One-Shot Case Study
X
O

Regression

History

Instrumentation

Internal

Interaction of
Selection and X

Table 8-1

-

-

+
+
+
+
+
+
+
+
+
?
?
Solomon Four-Group Design
R
O
X
O
R
O
O
R
X
O
R
O
Note: In the tables, a minus (-) indicates that the factor is not controlled, a plus (+) indicates that the factor is controlled,
a question mark (?) indicates a possible source of concern, and a blank indicates that the factor is not relevant
...
T
...
C
...

Copyright ©1963 by Houghton Mifflin Company
...


167
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Factorial Designs

A factorial experiment is one in which an equal number of observations is made of all
variable combinations
...
In
essence, the factorial design is one that has combined two or more completely randomized
designs into a single experiment
...
Suppose we return to our Ad illustration
but now assume that the researcher is interested in studying the effects of two variables of
interest: Ad Spokesperson type (at three levels – Male, Female, Cartoon Character ) and
Message Type (at two levels – Benefits Message and Humorous Message)
...
Note that each combination of MiSj occurs only once in the design
...

In the factorial experiment we can test for all main effects (i
...
, spokesperson types,
messages), and in this case, where we have replicated each combination, for the interaction of the
variables as well
...
That is, in market experimentation the researcher is typically interested in the
combination of controlled variables that leads to the best payoff in terms of preference, sales,
market share, cash flow, or some other measure of effectiveness
...
In using Latin-square designs the
researcher is usually assuming that interaction effects are negligible; in so doing, all main effects
can be estimated by this procedure
...
For example, the researcher may be interested
in the effect of shelf placement on the sale of energy drinks and be able to manipulate the
following variables for product placement:
A: shelf height—four levels—knee level, waist level, eye level, reach level
B: shelf facings—four levels—25%, 50%, 75%, and 100% of total width of gondola
C: shelf fullness—four levels—25%, 50%, 75%, and 100% of total height of gondola section

If the researcher were to run a full factorial experiment, with one replication only, there would be
(4)3 = 64 observations required
...


168
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Latin-Square Design—Energy Drink Shelf Placement Experiment

Table 8-3

Variable A—Shelf Height

B1

A1
A2
A3
A4

C1
C4
C3
C2

Variable B—Shelf Facing
B2
B3

C2
C1
C4
C3

B4

C3
C2
C1
C4

C4
C3
C2
C1

Table 8-3 shows one possible Latin-square design for this experiment
...
See
Edwards, 1968, pp
...

Randomized Block Design

Randomized-block designs, also known as treatments by blocks designs, represent a
frequently used experimental framework for dealing with multivariable classifications
...
For example, suppose that
our researcher were interested only in the effect of ad spokesperson effect on product
performance, but had designed the experiment so that more than one product was used in the
study
...
That is, each product
would be considered a test unit and each level of spokesperson would be tested in each product
...
Respondents could be randomly assigned to the
respective treatments, or even shown multiple treatments if measurement and learning effects
(distracters of internal validity) were ignored
...
Thus, if genuine treatment effects are present, this type of design will
be more likely to detect them than a single-variable classification in which the block effect
would become part of the error term
...
For example, if test units consist of human
subjects and the response variable is the number of correct identification of trademarks that are
shown on a projection screen (where such factors as length of exposure and clarity of focus are
varied), it may be that response is affected by the general intelligence level of the viewing
169
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

subject
...
We shall assume that the researcher is able to measure
each subject’s IQ
...
Roughly speaking,
the computational procedure is similar to a regression problem
...

Recapitulation
As noted earlier, the study of experimental design is basically the study of two things:
1
...

analysis of variance and covariance techniques for testing whether the various
treatment effects are significant
Many other kinds of design layouts, including fractional factorial designs, balanced incomplete
blocks, hierarchical designs, split-plot designs, and partially balanced incomplete blocks are
available
...

We have not covered the issue of how to select an experimental design
...
Such issues as the number of independent variables, the sources and
number of extraneous variables, the nature of the dependent variable, the number of subjects
available for participation in the experiment and other methodological issues will have a bearing
on what might be an appropriate design to use
...

The Panel as a Natural Experimental Design
The normal course of operation of a consumer panel generates a continuing set of natural
experimental data
...
Audience and deal panels provide similar response measurements
...
To illustrate their application, suppose that we have increased the price of
our product in selected territories
...
A cross-sectional analysis may be made by
comparing, for a given period after the increase, the purchase data for the territories in which the
price was raised with those in which no change was made [classical design (4)]
...
Such a study could employ either classical design (3) or (5)
...
A major difficulty, or course, is in sorting out the effect of the price increase
from the extraneous producers affecting purchases over time and among territories
...
Smith and Gerald S
...
History variables must therefore be analyzed carefully in using panel data
...
Consider, for example, a
market test of a general price increase
...
The remaining
territories automatically constitute the control group
...

The general advantages and limitations of these designs were discussed earlier
...

The Limitations of Continuous Panels
Although panels can provide highly useful marketing information that is difficult to
obtain by alternative research methods, there are some important limitations
...
To be most useful for drawing inferences about the population
being studied, the sample should be drawn by a random process
...
Panels require a great deal of effort to recruit and maintain
...
To reduce the bias introduced by such nonrandom attrition, replacements are typically
chosen from families with the same demographic and usage characteristics as those lost from
refusals and dropouts
...
Since the individual is undoubtedly conditioned to some extent by the fact that data on
purchases are reported, panel members may become atypical in their purchase behavior as a
result of being a part of a panel
...

Panel data may also be systematically biased through instrument effects
...
These are self-administered, structured questionnaires
...
However, in those cases where there is incomplete data possible biases are introduced
and the total amount of data is decreased
...
If the panel is administered carefully, the resulting data are
important additions to the information required for making sound marketing decisions
...
Smith and Gerald S
...
com survey software provides advanced survey capabilities and logic,
including experimental designs and data analysis
...
In addition, integrated panel
management capabilities make possible accurate tracking, profiling and monitoring of response
history, response timing and response progress for each respondent
...
These features provide fundamental tools for increasing control of
field experiments
...

Structuring Advanced Questionnaires
Advanced online questionnaires have at their foundation five essential building blocks
...
This functionality can include the development
of and presentation of experiments
...
Identify the populations to be sampled, along with experimental and control groups
2
...
Identify the organization of blocks of questions if needed for survey flow, branching or
experimentation
4
...
Define sample size quotas that are required for appropriate analyses
Note that we have not yet considered the important issues of measurement, including construct
validity, the wording of questions, or controlling survey content
...

Survey Realism, Interest and Respondent Involvement
Experimental design is all about presenting specific treatments to the respondent and measuring
the effects
...

Enhancing Survey Content: Graphics, Audio, Video and Interactive Experiences

Graphics increase interest and provide realism to a survey
...
A corporate logo at the top of the
survey may provide identity and authenticity to the questionnaire
...

Prior to online surveys, the use of audio creatives (songs, jingles, messages with varying
content, etc
...
Online technology facilitates the use of songs, commercials, legal arguments, and other
verbal, melody, or tone-based treatments in online concept tests and experiments
...
The increased adoption of high-speed
Internet continues to expand the type and amount of information that can be tested
...
Smith and Gerald S
...
Qualtrics accepts
browser interfacing programs that may be used to provide interactive Web Services content
(programmed using Flash, Java Script, PHP or Visual Basic)
...

As examples, these programs could control treatments for sophisticated experiments
using Latin-square or d-optimal Federov designs that present options in choice based conjoint
analysis
...
Similarly, a researcher could for each survey respondent randomly select, say, 5 open
ended text responses from other respondents who have already completed the survey, and then
ask the current respondent to rank the 5 extracted text statements in order of preference
...
The researcher is thereby unconstrained,
except in their own creativity
...
Piping generally refers to the movement of
information from one location to another within the survey
...
It may also include the
text of answer options selected, or those not selected
...

Travelocity customers receive a survey a few days after they return from a trip booked online
...
Instead, the
customer is addressed by name and is asked specific questions about their flight to Dallas on
Delta Airlines
...
Question
blocks are also customized for hotels and cruises booked for the trip
...


173
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Panel information linked from external databases can be used in a variety of ways
...

As a simple example of piping, consider pieces of information that are piped based on the
respondent’s answers to previous questions
...
The text “Hummer,” piped to an open ended text
question about dissatisfaction the car
...

Other uses of piping might include concept or logo tests where respondents select a
company logo or other concept-related graphic
...
At the advanced end of the spectrum, personal data for a given
respondent may be read from a CRM (customer relationship management) database, such as in
the Travelocity example cited above, and then piped into questions or answers within the
questionnaire
...

Figure 8-3

Piping Smart Text – Rich Text Editor Screen

174
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Creating Dynamic Surveys with Skip Logic, Branching and Looping and Control
Logic to control question order based on respondent answers can be easily added to online
surveys
...
We will begin with an introduction to simple skip logic and progress to
more advanced logic used in experimental designs
...
For example, if a respondent indicates that she is a nonconsumer of energy drinks, she is skipped past energy drink product and brand specific
questions, again entering the linear flow of survey questions that ask about general consumption
of casual beverages
...
Multiple instances of skip logic could
be entered where appropriate
...
Select “Add Skip Logic”
from the purple dropdown button to the left of the question
...
Multiple skip logic statements are supported for each question (Figure 8-4)
...
To Pipe Smart Text from Q14 to Q33, just go to Q33
and edit the question
...
The specific piped text is specified using the
menu options
...

Survey- and System-Wide Boolean Logic for Compound Branching

Branching is more sophisticated than linear skip logic and is used to skip questions based on a
complex sequence of conditions
...
Smith and Gerald S
...
She is “female,”
“between 18 and 34,” “married,” and has “one or more children
...
With branching, the intersection of these identifying characteristics is the point where we
change survey flow and question logic
...
This
approach enables a survey to branch based on (a) answers from multiple questions within the
survey; (b) outside variables nonspecific to the respondent, such as a randomly assigned
experimental group defined in either the panel information or by Web Services content; (c) panel
and database information about the respondent (such as CRM systems or past surveys the
respondent has completed)
...

Figure 8-5

Linear Skip Logic versus Boolean Logic and Compound

Looping with Piping

Looping with piping is a combination of features designed to reduce the length and complexity
of a survey building process when a standard set of evaluation questions are repeated over and
over
...

As an example, suppose that a respondent is asked to select from a list of 150 educational
programs, those that were used by his or her organization
...
Rather than building a linear survey of 600

176
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

questions (do you use program x? – yes, no; if yes, then evaluate, if no, ask about next program),
the researcher would build only 4 questions (check all that apply in the list of 150 programs and
the three evaluation questions)
...
The pipe and loop feature would loop through only those programs
selected, identifying them by name as the three evaluation questions were repeated
...

Note that looping based on a “Question Response” can be based on all choices presented to the
respondent in a previous question, all choices selected by the respondent, or those choices not
selected (for asking questions of the type: “why didn’t you consider this option in your
decision”), or those choices that appear or do not appear, where appearance occurs as a function
of some previously defined condition
...

In contrast to control by “Question Response”, the experimenter can use “Static” control with a
predefined set of loops and piping
...
In this case, the predefined product category and brand names “fast food,
McDonalds, Arbys, Subway” could be piped and merged into the sequence of three questions
that is automatically repeated for each of the three fast food brands evaluated
...
This predefined list is defined in the Block
Options / Loop and Merge dialogue box as shown in figure 8-6
...

Figure 8-6

Block Loop and Merge Static Text (Researcher Provided)

With increasing time demands on potential respondents, they often review questionnaires to
estimate the time and effort required
...
Smith and Gerald S
...

Carry Answers Forward

Closely related to the streamlining effects of loop and pipe feature is the ability to carry forward
any or all respondent answer choices from one question to a later question
...

All possible answers or only the k of n items that were selected can be moved to a later
question to ask further questions about those items
...
The order of item presentation can be maintained or
randomized
...
This text response
can then be carried forward to gain additional information
...
Quotas work by testing the size of
the sample groups (defined by answer categories within a given question) against target numbers
(Figure 8-7)
...
Given that many studies pay
respondents for completed surveys, quotas, if used early in the questionnaire can screen and
eliminate unneeded respondents, thereby reducing total research costs
...

Figure 8-7

Sample Size Quota Setting

This section has provided a brief overview of the embedded logic and complex information
flows that can be adapted to meet the rigors of an experimental design
...
In addition, experiments
and statistically sound quota sampling are facilitated in dynamic sequences of questions that are
displayed based on a respondent’s answers and/or match to a group or quota
...


178
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Using Survey Flow to Create Experiments
Qualtrics contains powerful, but easy to use tools for customizing question and block
presentation order in experimental designs
...

Question Blocks

A question block is a cohesive group of questions that receive similar treatment
...
More
advanced questionnaires may split questions into multiple question blocks based on topic or
logical flow
...
Questions
within blocks may be arranged in a fixed order, random order, or partial random order that holds
some questions in a fixed position, while randomizing the order of other questions
...

Any element in the survey flow can be dragged and dropped into any location in the flow
...

Randomizing Questions and Blocks of Questions

Randomization is used when it is not important to maintain the question order or context
of the questions
...
For example, context often requires that questions be
presented in a given sequential order, but blocks of questions (each block, for example,
containing questions about different brands) can be randomized
...


179
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 8-8

The Survey Flow Chart

Once the survey building has been completed and question blocks are created, the initial
survey flow is created
...
The survey
follows the main flow line on the left, and presents questions in the order shown from top to
bottom
...
Survey flow
Elements may be added at any point in the flow
...
The two
survey flow elements most often used are “Block" (question blocks), and "Branch
...

Figure 8-9

Survey Flow Elements

Adding Block and Branch Elements to the Survey Flow
Block
Blocks of questions, once created, are automatically included in the questionnaire flow
...
Once a block is inserted, it
will appear as a survey flow element on its own secondary line
...


180
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Branch Logic
Branches function differently than Blocks
...
Those not meeting the condition proceed to the next condition in the
survey flow
...
The logic section can also be
triggered by choices that were Selected, Not Selected, Displayed or Not Displayed
...
Displayed and Not Displayed options
are useful when text is piped or answers are carried forward from one question to another
...
Options are
specific to the question type
...

Figure 8-10

Multiple Then Branch If Statements

Note: Because experimental designs often require extensive logic sequences, it is worth noting
that within the chain of logic statements, quotas have first priority, followed by the "And If"
clause, which has a higher priority than the "OR IF" clause (i
...
with a chain of clauses such as 1
AND 2 AND 3 OR 4, the system will read it as (1 AND 2 AND 3) (OR 4), meaning IF 1 AND 2
AND 3 are all fulfilled, OR if 4 is fulfilled, then the respondent will be branched
...
Generally, embedded data will be
uploaded as part of panel data
...

For example, if we were to access the person‘s home state as an embedded data field in a
panel mailing list, we would specify the location, variable, logic and condition as:
"If: Embedded Data : Home State Equals Colorado
...

181
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Randomization of Answer Choices, Branches, Question Blocks and Surveys
Randomization techniques are used to control answer presentation order bias, as well as the
effects of extraneous variables not controlled by an experimental design
...

The randomization of choice options is the most basic type of randomization, and deserves little
special attention other than to suggest that answer choice sets without a fixed order and matrix
type questions (displaying lists of questions, each evaluated on the same scale) are best presented
in random order
...

Figure 8-9

Block Randomizer Screen

Randomization of treatment blocks works by inserting the element "Block Randomizer" into the
survey flow
...
The randomized blocks will appear in the
survey in the place where you insert the block randomizer
...
To select a subset of blocks, make sure all blocks to be randomized have
182
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

been added into the Block Randomizer
...
You may set it to show all of the blocks in the
randomizer, or a subset of those blocks (i
...
, If 4 blocks are added to the randomizer and each
respondent is to randomly respond to two of those blocks, put a "2" in the box
...
In such cases, the questionnaires are prepared and randomly
assigned to the respondents
...
The java script would be
similar to the one that follows to randomly assign respondents to a survey
...
observe(window, 'load', function() {

// Array structure - as many links as you need
var seedcount = 7
var index = 3;
var linkArray = new Array(index);
// Load array with available survey links here:
linkArray[0] = "http:// new
...
com/SE?SID=SV_6LR6QjjYiHkSe21&SVID=Prod";
linkArray[1] = "http:// new
...
com/SE?SID=SV_6LR6QjjYiHkSe22&SVID=Prod";
linkArray[2] = "http:// new
...
com/SE?SID=SV_6LR6QjjYiHkSe23&SVID=Prod";
// Generate random number that will correspond to an index in the above array
...
getSeconds();
randNum = Math
...
random(seed))
for (loopcount = 0; loopcount < seedcount; loopcount++) {
randNum += Math
...
random(randNum * seed));
}
randNum = Math
...
random(randNum)) % index;
// Use one of the two statements below by changing the comment
// Redirect user to appropriate random survey page in the current window
window
...
replace(linkArray[randNum]);
// Or open the appropriate random survey page in a new window
// window
...
Smith and Gerald S
...

Consider the case of a sales manager who wishes to determine the effects of varying amounts of
sales effort on product sales of a nationally distributed brand
...
However the capabilities of online surveys software like Qualtrics
...

Experimentation represents another way of information gathering and should be
approached in terms of its cost versus value (compared with other techniques) for decisionmaking purposes, with regard to both present and anticipated problems
...


SUMMARY
In this chapter our primary objectives were (1) to introduce the conceptual bases
underlying marketing experimentation, and (2) to discuss basic models of experimental design
...

The first section of the chapter covered the nature of experimentation
...
We next turned to the
potential sources of invalidity, internal and external, associated with an experiment
...

We next described some of the major classical and statistical experimental designs
...

The Qualtrics research suite was then presented as a powerful and integrative tool for
researchers who need to conduct advanced field studies, including experimentation
...

We have also discussed advanced logic and flow capabilities that support the creation of
experiments
...


184
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

REFERENCES
Banks, S
...
Experimentation in Marketing
...
(1998)
...

Brown, S
...
and Melamed, L
...
(1990)
...
Thousand Oaks,
CA: Sage Publications,Inc
...
T
...
C
...
Experimental and Quasi-Experimental Designs for
Research
...
B
...
Experimental Methodology
...
D
...
T
...
Quasi-Experimentation: Design and Analysis Issues for
Field Settings
...

Edwards, A
...
(1968)
...
New York: Holt,
Rinehart & Winston, Third Edition
Field, A
...
(2003)
...
Thousand Oaks, CA:
Sage Publications
Perdue, B
...
and Summers, J
...
(1986)
...
and Rosnow, R
...
(1975)
...
New York: John Wiley & Sons
Rosenthal, R
...
L
...
Essentials of Behavioral Research: Methods and Data
Analysis
...
G
...
“Demand Artifacts in Laboratory Experiments in Consumer Research,”
Journal of Consumer Research, 1 (March)
Shadish, W
...
D
...
T
...
Experimental and Quasi-Experimental
Designs for Causal Inference
...


185
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Appendix 8
...
g
...
e
...

With the goal of creating stimuli to test the hypotheses, a well known automotive brand
(CADILLAC) was chosen as the senior mark
...
A
fictitious junior mark (CADILLAC brand notebook computers) as then created, along with an
advertisement that described an upscale, luxurious notebook computer that could be offered by a firm
attempting to infringe on the CADILLAC trademark
...
”
Cadillac brand notebook computer! This luxury notebook has an upgraded keyboard for better
feel and absolutely silent typing
...
Matching dark
mahogany accents are found on the wireless mouse and the leather carrying case
...
Price available upon request
...

To test this hypothesis, three stimuli were created
...
The advertisement emphasized computing power and did not attempt to create an upscale
positioning with luxurious product attributes
...

Luxurious product attributes such as a titanium case were described in support of an upscale positioning
...

The third stimulus (see Figure 3) informed participants of the launch of an upscale notebook
computer as a brand extension of a well-known automotive brand (MERCEDES-BENZ)
...
By
comparing viewers of this stimulus with viewers of the two control stimuli, H1 can be tested
...
The experimental method was used
to provide a rigorous test of the causality claimed in H1
...
The stimulus
was shown in full-color, and the time of exposure was self-paced
...
They were then shown the advertisement for the CADILLAC notebook
computer and asked to complete the source-confusion battery
...
”
1
Source: Lee, T
...
, Derosia, E
...
and Christensen, G
...
, Sophistication, Bridging The Gap, And The
Likelihood Of Confusion: An Empirical And Theoretical Analysis, The Trademark Reporter, Vol
...

4, July-August 2008
...


186
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 1 (Condition 1: Control)

Figure 2 (Condition 2 (Control) Figure 3 Condition 3 (Treatment)

The following is the flow chart shows the above three group experimental design with control

187
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Case 2:

Determining Price Sensitivity

A major investment firm is interested in determining the price sensitivity of fund managers to
specially prepared research studies about selected new equity offerings
...
This analysis is followed by an experimental design (unrelated to the conjoint
analysis) that measures the fund manager’s sensitivity to the report price given their current payment
expectations for a single research study that is to be prepared exclusively for their firm
...

The Experimental design in this example would split respondents based on the price they
currently expect to pay for a single exclusive research report ($20k, $30k, $40k, $50k)
...
The prices tested will vary as shown in Table 1
...
2K
10@$81
...
8K
100@$20
...
6K
10@$28
...
4K
100@$7
...
4K
10@$43
...
6K
100@$10
...
8K
10@$122
...
2K
100@$30
...
2K
10@$57
...
8K
100@$14
...
4K
10@$163
...
6K
100@$40
...
The flow next branches based on the
respondent’s stated acceptable price, to a matrix evaluating exclusivity x volume discount
...
Finally, Branch elements direct the respondent to the appropriate pricing sensitivity section
...
Smith and Gerald S
...
Smith and Gerald S
...
Smith and Gerald S
...
Marketing decisions ultimately hinge,
in whole or in part, on a prediction of the behavior of consumers, industrial users,
marketing intermediaries, competitors, and, at times, the government
...

We now consider the most frequently studied variables used to predict behavior:
demographics, attitudes, intentions, satisfaction and loyalty
...

Behavioral Correlates
Behavioral correlates address the question of what past behaviors or intended
behaviors predict future behavior
...
Each of us relies
heavily upon our knowledge of others’ past behavior in our everyday relationships with
our family, friends, and associates
...
In economics applications, we examine past
trends, seasonal averages, and cyclical patterns for forecasting
...
Recognizing that the degree of stability is
sometimes difficult to determine, and that we do not always understand the underlying
causal relationships, we nonetheless must believe that there is some continuity and
stability in the behavior patterns of people
...

Information about past behavior toward products is often classified into three
categories: acquisition, use, and possession
...
This data produces detail useful in understanding product
consumption patterns
...
Table 9
...
Smith and Gerald S
...
Often, such information comes from secondary sources, as previously
discussed
...
What more logical method of predicting the future behavior of
respondents could be used than determining their intentions? After all, intentions are selfpredictions of behavior, and thus, if obtained from people whose behavior we want to
predict, would seemingly be the most direct and reliable method of prediction
...
However,
consideration of our own experiences in terms of what we have planned to do vis-à-vis
what we have actually done later should raise some questions concerning the reliability of
intentions as a predictive tool
...

Past Behaviors: A Study of Tomato Juice Usage Patterns

TABLE 9
...

192
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The results of a study of expected and purchase rates of a few products and
services are shown in Table 9
...
Researchers collected intentions data from a consumer
panel sample using a 0 to 10 scale to measure purchase probabilities
...
A 10 was defined as “absolutely certain of
buying” and a 0 as “absolutely no chance of buying
...

Intentions Fulfilled and Not Fulfilled During a 60-Day Period
Intentions-Based
Expected Purchase
Product/Service
Rate %
Purchase %
Difference
Ride local public transportation
22
...
7
-0
...
4
7
...
2
17
...
6
-2
...
2
3
...
5
Take a trip on cruise ship
14
...
1
-0
...
2

Expected purchase rates were calculated as the average purchase probability for
each item
...

Intentions to buy are often conditioned by judgments and expectations of future
events or future situations, as well as by our past experiences
...
Since each of these is subject to
variation, it seems plausible to suppose that the intender views them as such and that his
or her stated intention is based on a subjective probability of purchase
...

Past experiences can often be inferred from a customer’s level of satisfaction
...
Customer satisfaction relates to
intentions as it can significantly affect the repurchase decision of consumers
...
We now need to examine the nonbehavioral correlates
that are useful for predicting their future behavior
...
The Radio Corporation of America (RCA), when introducing color television
in the 1950s, was very much interested in the age, income, educational, and occupational
composition of the market
...
These judgments were subsequently confirmed by a study of early
purchasers of color sets
...
Smith and Gerald S
...
1 Measuring Intentions

194
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

decisions, since an association was found to exist between families with these
characteristics and the purchase of color televisions sets
...
While socioeconomic
characteristics are by far the most widely used bases for classification of consumers,
other bases exist
...
It may be interesting to know, for example, that owners of SUVs show different
personality traits than owners of other vehicles; such knowledge will be useful in
marketing automobiles, however, only if it can be used to develop and evaluate appeals
for each type of buyer
...

In general, the identification of consumer segments is useful in marketing so long
as the following four statements apply:
1
...

2
...
There is homogeneity within and heterogeneity
between segments
...
Operational: There is a cost effective means of reaching the targeted market
segment
4
...

Two commonly used and widely accepted classifications of consumers are by
stage of the life cycle and by lifestyle
...
Young unmarrieds
2
...
Young marrieds, with children, youngest child under six
4
...
Older marrieds, with children maintaining separate households
6
...
See Wells and Gubar (1966)
and Wagner and Hanna (1983) for more detailed explanations of the life-cycle concept
and marketing research
...
Other, less obvious
relationships exist as well
...
A second stage of furniture buying
195
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

takes place when children begin to date and have parties at home
...

Lifestyle has a close association with membership in a social class
...

These differences tend to be expressed through the products bought and stores
patronized, as well as the area in which one lives, club membership, religious affiliation,
and other means
...
When used with personality traits, lifestyle
variables form the basis of psychographic classification, as illustrated in Exhibit 9
...
An
illustration of psychographic questions is shown in Table 9
...

As an example, let us consider the life styles of the Harley Owners Group (HOG)
...

EXHIBIT 9
...
Such schemas
represent interesting demographic and product markets, and provide a much more colorful
description of the group as a whole as well as the diversity within
...
They like riding
long distances, use their bike both for touring and everyday transportation, are more
interested in the comfort of their motorcycle than its speed, prefer riding with a
passenger, and wear a helmet
...
They are less ambitious than
others, and are distinctively unattracted by social gatherings and danger
...
The Vanilla Dream Riders are more interested in the dream of
motorcycling than in motorcycling itself, and are otherwise just plain vanilla—a
relatively undistinguished group
...
You see the Dream Riders taking riding on short trips
around town (often by themselves), wearing a helmet, and riding a stock bike
...
Their motorcycle is merely a possession, having no
real place as a “family member
...

The Hard Core segment is on the fringe of society, and identifies with the stereotypical
biker subculture
...
Virtually all are blue-collar workers
...
Smith and Gerald S
...
” Note, however, that they
still only “slightly agree” that these lifestyles describe them well
...
”
Hog Heaven finds great psychic and spiritual satisfaction in owning and riding a Harley
...
They have many motorcycle friends, and when group riding,
they feel the group becomes “one
...
”
This segment is distinctively mechanically inclined, likes to work on their
motorcycles, but spends little on accessories
...
They are less likely than others to believe in a life
after death, but often think about how short life really is
...

They include the highest percentage of married riders, but otherwise are typical of
Harley owners in most demographic characteristics
...

Zen Riders are more impulsive and are willing to take chances and to run risks
...
They like to party, and have
trouble relaxing in everyday life, but they are “modern” husbands and are opposed
to legalizing marijuana
...
They “ride to live and live to ride
...
They simply love riding
...
They find riding to be a
magical experience, and motorcycling is a total lifestyle to them
...
They agree with an “eat, drink, and be merry” premise, like to create a stir, like
danger, and get lots of satisfaction from their hobbies
...


Life style segmentation offers significant implications for marketing and advertising to
these segments in terms of not only better communicating with each group, but in
creating the product configuration and accessories that appeal to each group
...
Thus, the Warnerian classification results in
six classes, ranging from the UU (upper upper) down through the LL (lower lower)
...


197
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

EXHIBIT 9
...
Smith and Gerald S
...
The media to which the managerial-professional, upper-middleclass family is exposed, and the appeals to which it responds, are also likely to be closer
to those of other managerial-professional families than to those of the blue-collar family
...
The marketer can then direct the product at this lifestyle group
...
These values fit well into the qualitative research discussed in Chapter
6, (means-end laddering to produce hierarchical value maps and ZMET analysis):
Self-respect
Security
Warm relationships with others
Sense of accomplishment
Self-fulfillment
Being well-respected
Sense of belonging
Fun and enjoyment in life
Classification of consumers is vital if we are to learn more about consumer
behavior and utilize this information to develop more efficient marketing techniques
...
Given the new software available that simplifies
using such advanced techniques as neural networks, latent-class models, fuzzy or
overlapping clustering, and even occasion based segmentation, there will be a tendency to
use a backhoe when a shovel is more appropriate
...
Advanced statistical techniques for consumer classification purposes are discussed
in later chapters
...
Comparable characteristics of these firms include sales volume, number
of employees, and the type of products manufactured or handled
...
” This is especially so when making advertising budget and media
allocation decisions, where consumers’ choices are strongly affected by their levels of
awareness and extent of knowledge of the product and its attributes
...
One question sought knowledge
of interest rate differences:

199
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Attitudes and Opinions
Investigators in the behavioral science fields of psychology, sociology, and
political science have made extensive studies of attitudes and opinions over a wide range
of subject areas
...
As such, it has been a natural consequence that marketing
research has adopted, adapted and applied many techniques, including attitude-opinion
studies to obtain information applicable to the solution of marketing problems
...
A commonly drawn distinction has been to
view an attitude as a predisposition to act in a certain way, and an opinion as a
verbalization of an attitude
...

When used to predict actions that the respondent will take, this distinction
between attitude and opinion rapidly becomes blurred
...
We shall therefore use the terms interchangeably
...
In either form, researchers encounter problems that are more
severe than those involved in obtaining any other type of descriptive information
discussed
...
Its greatest use has been in the areas of product design (including
packaging and branding) and advertising
...
In fact,
attitudes and opinions are central in customer satisfaction studies
...
Expectancy value
models use attitudes to predict behavioral intention (intention to try, purchase,
recommend, or re-purchase a product or service)
...

The Expectancy value model uses attitudes and beliefs in a mathematical formulation that
is read as follows:

200
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Behavior (purchase of brand X) is approximated by Behavioral Intention (intention to
purchase brand X), which in turn is approximated by the Overall Attitude toward brand X
...

Normative beliefs are added to this compensatory evaluative process by considering the
norms surrounding an attitude ( My friends insist I should buy an environmentally friendly
car that gets 50 mpg), and applying weights reflecting the buyer’s motivation to comply with
the norm (Will they praise me for buying a 50 mpg car? – or Do I really care about getting
50 mpg?)
...

Ao = Overall Attitude toward the object (brand, product, service or company)
w1, w2 = Weights that indicate the relative influence of the overall attitude toward the object
and the normative influence to purchase the product
ai * bi = the sum of individual attitudes toward the object, weighted by importance
...
The sum is taken over the k attributes that are defined as salient in the purchase
decision
...


Attitude (ai*bi)
ai is the affective (liking) component of the evaluation of attribute i
...
The evaluation would use a five or seven point scale with endpoints ranging
from “Poor” to “Excellent”, or “Not at all Desirable” to “Very Desirable”
...
1
...
This is sometimes
measured as the probability of attribute i being associated with brand X
...
Using the context of
purchasing a Toyota Prius, the attribute would be “Gets 50 miles per gallon” could be rated on a
seven point scale with endpoints labeled "Very Unlikely" and "Very Likely", or as in the
example below, is measured as “Not at all Important” to “Very Important”
...
1
...
Smith and Gerald S
...
1

Expectancy Value Importance-Performance Rating Scales

nbi * mci = The overall normative component of the decision process
...
2
...
2

Normative Rating Scales

The normative component is often considered to be absorbed in the overall attitude
component and depending on the application, is often ignored
...
Smith and Gerald S
...
This compensatory type of decision model results in a
powerful approach to measuring customer attitudes and predicting customer behavior, and is
well used in research
...

Measuring Satisfaction
Customer satisfaction is the most common of all marketing surveys and is part of
the “big three” research studies in marketing that include market segmentation and
concept testing
...
51)
...
But measuring satisfaction is not measuring loyalty
...
Overall, how satisfied are you with (brand name)?
2
...
Do you intend to repurchase (brand name)?

According to William Neal (2000), these questions are usually measuring the
same thing—satisfaction with the product or service
...
Customer satisfaction can be measured by
traditional surveys, by using comment cards, and for business-to-business situations by
field or call reports, to name just some of the methods (Cochran, 2001)
...
4)
...
Consequently, the marketing
manager should be as well informed as possible about both the nature of their relevant
attitudes and opinions and the intensity with which they are held
...
Findings from these surveys are reported back to the dealer, who then acts on
them
...
This survey examines the quality of the product and whether new owners are
satisfied with the performance, fit, and finish of their new vehicle
...
We will first introduce the theoretical and
methodological underpinnings of satisfaction research by first defining the concept of
customer satisfaction and how satisfaction is used in business
...
Finally, the components of a
satisfaction survey are presented, along with sample satisfaction survey questions
...
com survey library contains sample satisfaction questionnaires, questions and
scales
...

203
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Exhibit 9
...
)
Knowledge and Involvement with product and the purchase process
Awareness of other brands
Reasons for original product purchase (selection reasons)
Primary benefits sought from the product
Product Evaluation
Attribute evaluation matrix: (quality, price, trust, importance, performance, value)
Perceived benefit associations matrix
Importance, performance
Identification of primary benefits sought
Comparison to other brands (better, worse)
What is the best thing about the brand, what could be done better
Message and Package Evaluation
Packaging size, design
Advertising Promise, message fulfillment evaluation
Value Analysis
Expectation of price
Expectation of relative price (full price, on sale)
Current price paid
Satisfaction Measurements
Overall Satisfaction
Reasons for Satisfaction Evaluation
Satisfaction with attributes, features, benefits
Satisfaction with use
Expected and Ideal Satisfaction-Performance Measures
Likelihood of recommending
Likelihood of repurchasing

What Is Customer Satisfaction?
Customer satisfaction measures indicate how well a company’s products or services meet
or exceed customer expectations
...
Customer satisfaction
measures will tap the customer’s lifetime of product and service experience
...
Smith and Gerald S
...
Customer satisfaction measures are critical to any company, be it consumer or
B2B, because customer satisfaction is a strong predictor of customer retention, customer loyalty
and product repurchase
...

Three very different approaches may each produce meaningful and useful results
that are appropriate for specific situations, uses and needs:
Post Purchase Evaluations reflect the satisfaction of the individual customer at
the time of product or service delivery (or shortly afterwards)
...

Continuous Satisfaction Tracking is often part of a management initiative to
assure quality is at high levels over time, and can involve post purchase
evaluations or a succession of periodic satisfaction surveys conducted on a regular
basis (daily, quarterly or monthly basis)
...
Satisfaction tracking surveys are
...
Satisfaction surveys typically require multiple questions
that address satisfaction with different dimensions of the product or service concept
...
Satisfaction measurement is like peeling away layers of an onion-each
layer reveals yet another deeper layer, closer to the core
...
Ulwick (2005) advocates the combination of satisfaction and
importance to identify unfulfilled opportunities in the marketplace
...


205
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Customer Satisfaction Survey Measures

Customer satisfaction surveys will often include several different multiple measures of
satisfaction, including:
Overall measures of customer satisfaction
Affective measures of customer satisfaction
Cognitive measures of customer satisfaction
Behavioral measures of customer satisfaction
Expectancy value measures of customer satisfaction
We will now explain each in detail
...

Satisfaction Measurement: Overall Measures of Satisfaction
Like when we have a great food experience at a favorite local restaurant, elevated levels
of customer satisfaction usually leads to customer loyalty and product repurchase
...
Satisfaction measurement questions
typically include items like those found in Exhibit 9
...
An overall satisfaction measure (emotional):
Overall, how satisfied are you with “Yoni fresh yogurt”?
This question reflects the overall opinion of a consumer’s satisfaction experience
...
There is no experience base for such a measure of satisfaction
...
A loyalty measure (affective, behavioral):
Would you recommend “Yoni” to your family and friends?
3
...
The psychological concepts of attitude
and satisfaction may both be defined as the evaluation of an object and the individual’s
relationship to it
...

4
...

Additional post-experience actions might reflect a heightened level of product involvement
206
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

that in turn results in increased product or information search activity, reduced trial of
alternative products, and even changes in preferences for shopping locations and choice
behavior
...
3, customer satisfaction is diagrammatically shown to be
influenced by perceived quality of product and service attributes, associated product features and
benefits, and is moderated by the customer’s expectations regarding the product or service
...

Affective Measures of Customer Satisfaction
Attitudes toward a product can be developed as a result of product information or any
experience with the product, whether perceived or real
...
Again, it may be meaningful to measure
attitudes towards a product or service that a consumer has never used, but it is not meaningful to
measure satisfaction when a product or service has not been used
...
3

Building a Customer Satisfaction Survey

207
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Exhibit 9
...
Smith and Gerald S
...
Smith and Gerald S
...
Smith and Gerald S
...
Smith and Gerald S
...
Cognitive responses are often specific to the situation for which the product was
purchased and specific to the consumer’s intended use of the product, regardless if that use is
correct or incorrect
...
" When phrased in behavioral response terms for a future or hypothetical behavior, consumers
indicate that “purchasing this product would be a good choice” or “I would be glad to purchase
this product
...
e
...

Expectations Measures of Customer Satisfaction
As might be expected, many different approaches to measuring satisfaction exist in the
consumer behavior literature
...
SERVQUAL has become the standard for
service quality measurement, but it should be recognized that satisfaction dimensions seem to
vary depending on the application (high tech vs
...
leisure services, etc
...

Figure 9
...
Smith and Gerald S
...
This “satisfaction gap" involves measuring both perception of performance and
expectation of performance along specific product or service attributes dimensions
...

Again we see that customer satisfaction is largely a reflection of both expectations and
experiences that the customer has with a product or service
...
For example, when we make major purchases, we research
the product or service, collecting information from the advertising, salespersons, and word-ofmouth from friends and associates
...

Customer Expectations that Influence Satisfaction
Customers hold both explicit and implicit performance expectations for attributes,
features and benefits of products and services
...

Explicit expectations are mental targets for product performance, such as well identified
performance standards
...

Implicit expectations represent the norms of performance that reflect accepted standards
established by business in general, other companies, industries, and even cultures
...
Performance measures for each application are unique, though general
expectations relate to quality of outcome and may include those researched by Parasuraman or
others such as: accessibility, customization, dependability, timeliness, and accuracy, tangible
cues which augment the application, options, cutting edge technology, flexibility, and user
friendly interfaces
...

Dynamic performance expectations are about how the product or service evolves over
time and includes the changes in support and product or service enhancement needed to meet
future business or use environments
...

213
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Technological expectations focus on the evolving state of the product category
...
Mobile service providers, in an effort to deal
with the desire to switch to new technology phones, market rate plans with high cancellation
penalties for switching providers, but with liberal upgrade plans for the phones they offer
...
These highly involving products are not just feature
based, but have expectations that enhance perceptions of status, ego, self-image, and can even
invoke emotions of isolation and fear when the product is not available
...
Person to person relationships are increasingly important, especially where
products require support for proper use and functioning
...

Situational Expectations
In building a customer satisfaction survey, it is also helpful to consider reasons why prepurchase expectations or post-purchase satisfaction may or may not be fulfilled or even
measurable
...
Such expectations are not only
difficult to evaluate and understand, but may change over time
...

When fulfilled, expectations result in customer satisfaction (or when expectations are not
fulfilled, result in dissatisfaction and complaining behavior)
...
The research study may
also benefit from consideration of expectations related to perceived quality and value
...
Smith and Gerald S
...

Customer experiences that result in attributions of quality are the single greatest predictor of
customer satisfaction
...
Perceived value is measured in many ways including overall
evaluation of value, expectations of price that would be paid, and more rigorous methodologies
including the Van Westendorp pricing analysis and conjoint analysis
...

The consumer behavior literature has long shown that price is a primary indicator of
quality when other attributes and benefits are relatively unknown
...
This may
reflect a decision simplification strategy that makes shopping or purchase decisions easier, or
may reflect increased customer loyalty
...
Customer
satisfaction is a major predictor of repurchase, but is strongly influenced by explicit performance
evaluations of product performance, quality, and value
...
For a common
measure of loyalty might be the sum of the following three questions measured on an agreement
scale (Strongly agree – Strongly disagree):
1
...
How likely are you to continue to choose/repurchase [brand]?
3
...
Switching may occur because of out of stock
situations, inconvenience of going to a store that has the preferred brand, or the lack of
commitment to the brand
...
3)
...
3
Brand Loyalty Measures
Construct
Measure
Confidence in the brand
Integrity of the brand

Pride in the brand
Passion for the brand

[Brand] is a name I can always trust
...

[Brand] always treats me fairly
...

[Brand] always treats me with respect
...

I can't imagine a world without [brand]
...
Smith and Gerald S
...
5

Gallup Customer Loyalty Example

The application of these loyalty scales enables the differentiation of brands, companies
and even industries according to how well they are “engaging” their customers (Figure 9
...
hat
This approach to measuring customer loyalty opens the door for the creation of
performance benchmarks and for management decision making
...
Customer loyalty is created by front-line employees who interact
with the customer
...
Dealing with expectations in a timely and effective
manner increases customer loyalty and retention
...
As we have discussed, these expectations are based on the consumer’s
reservoir of affective, cognitive and behavioral experiences
...

This can include overall expectations or expectations for the individual product attributes
most important in the purchase decision or experience
...

3) Fulfillment of Expectations: the expected level of performance vs
...

This is “Predictive Fulfillment” and is a respondent specific index of the performance level
necessary to satisfy
...
If a
product/service is not used as often as expected, the result may not be as satisfying as
anticipated
...

216
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

SUMMARY
Customer satisfaction is the most important element of business success and is a key
measure of fulfillment of business strategies, including those involving segmentation and
concept development
...
The different satisfaction survey measures discussed were presented as components of
a satisfaction survey, along with sample satisfaction survey questions
...
These templates will assist you in further understanding what
measures should be included in your satisfaction research and why those measures are necessary
and of value
...
(2001, November)
...
Quality
Digest, 45–50
...
(2001, September 24)
...
Marketing News, 35, 1ff
...
(2002, May 13)
...
Marketing News, 36, 15–16
...
(1983)
...
New York: Warner Books
Neal, W
...
(2000, June 5)
...

Marketing News, 34, 19
...
Zeithaml, and L
...
"SERVQUAL: A Multiple-Item Scale for
Measuring Customer Perceptions of Service Quality," Journal of Retailing, Spring 1988,
pp
...

Ulwick, A
...
, & Hanna, S
...
The effectiveness of family life cycle
variables in consumer expenditures research
...

Wells, W
...
, & Gubar, G
...
Life cycle concept in marketing research
...


217
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

This page is intentionally blank

218
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 10
GENERAL CONCEPTS OF MEASUREMENT
AND SCALING
Questions focus on the problem we are trying to solve, while answers are more closely associated with the
measurement scale we are using to achieve our analysis
...
Regardless of the
research topic, useful data is obtained only when the researcher exercises care in making
procedural decisions such as:
1
...

3
...


Defining what is to be measured
Deciding how to make the measurements
Deciding how to conduct the measuring operations
Deciding how to analyze the resulting data

Definitions and decisions play a significant role in scientific inquiry, especially in
marketing research and the behavioral sciences
...
Increasingly, behavioral scientists are paying greater attention to defining
the concepts measured in their specific disciplines, and refining operational definitions that
specify how to measure and quantify the variables defining the concepts
...
This section serves as useful background for the
discussion of statistical techniques covered in later chapters
...

The overall quality of a research project depends not only on the appropriateness and
adequacy of its research design and sampling techniques, but also on the measurement
procedures used
...

Definitions in Marketing Measurement
Marketers measure marketing program success as increased brand awareness, ad
awareness, ratings of brand likeability and uniqueness, new product concept ratings and purchase
intent, and customer satisfaction (Morgan, 2003)
...

Models are representations of reality and therefore raise the fundamental question of how
well each model confidently represents reality on all significant issues
...
Validity refers to a model’s accuracy in
describing and predicting reality, whereas utility refers to the value it adds to the making of
decisions
...

Model quality also depends on completeness and validity, two drivers of model accuracy
...
Smith and Gerald S
...

Clearly, managers will probably benefit from models that are simple enough to
understand and deal with
...
The required sophistication of
a model to be used depends on the model’s purpose
...
Models should be used only if they can help us
arrive at results faster, with less expense, or with more validity
...
To better understand this, we
must briefly study the building blocks of measurement theory: concepts, constructs, variables,
operational definitions, and propositions
...

“Mass”, “strength”, and “love” are all concepts, as are “advertising effectiveness”, “consumer
attitude”, and “price elasticity”
...
For example, the construct “attitude”
may be defined as “a learned tendency to respond in a consistent manner with respect to a given
object of orientation
...
Variables are constructs
in measured and quantified form
...
e
...

Operational Definitions
We can talk about “consumer attitudes” as if we know what it means, but the term makes
little sense until we define it in a specific, measurable way
...
It is a
set of instructions defining how we are going to treat a variable
...
For example, the concepts of “attitude” and “importance”, which have
been summed as the multiplicative product of a series of attributes like cleaning ability, fresh
smell, etc
...
Smith and Gerald S
...
It is not enough to simply
state that the concept “sales” is a function of the concept “advertising”, such that S = f (Adv)
...

Integration into a Systematic Model
A proposition is quite similar to a model
...
When concepts,
constructs, variables and propositions are integrated into a model for a research plan, we should
conceptually ask the following questions:
Are concepts and propositions specified?
Are the concepts relevant to solving the problem at hand?
Are the principal parts of the concept clearly defined?
Is there consensus as to which concepts are relevant in explaining the problem?
Are the concepts properly defined and labeled?
Is the concept specific enough to be operationally reliable and valid?
Do clear assumptions made in the model link the concepts?
Are the limitations of the model stated?
Can the model explain and predict?
Can the model provide results for managerial decision making?
Can the model be readily quantified?
Are the outcomes of the model supported by common sense?

If the model does not meet the relevant criteria, it probably should be revised
...

INACCURACIES IN MEASUREMENT
Before delving into measurement scales and question types, it is helpful to remember that
measurements in marketing research are rarely “exact
...
A portion of this variation among individual scores may represent
true differences in what is being measured, while other variation may be error in measurement
...
Smith and Gerald S
...
For instance, a company wanting to measure
attitudes toward a possible new brand name and trademark would like to feel confident that
measurement differences concerning the proposed names represent only the individuals’
differences in this attitude
...
Measurements are often affected by characteristics of individual respondents such as
intelligence, education level, and personality attributes
...
Unfortunately, this type of situation cannot be easily
controlled unless the investigator knows all relevant characteristics of the population members
such that control can be introduced through the sampling process
...
Measurement error
has a constant (systematic) dimension and a random (variable) dimension
...
Systematic error is present because of a flaw in the
measurement instrument or the research or sampling design
...
These two subtypes
of measurement error affect the validity and reliability of measurement, topics that are discussed
in the later part of this chapter
...

MEASUREMENT CONCEPTS
Measurement can be defined as a way of assigning symbols to represent the properties of
persons, objects, events, or states
...
Another way of looking at this is that measurement is
“the assignment of numbers to objects to represent amounts or degrees of a property possessed
by all of the objects” (Torgerson, 1958, p
...
If a characteristic, property, or behavior is to be
represented by numbers, a one-to-one correspondence between the number system used and the
various quantities (degrees) of that being measured must exist
...

2
...


Order
...

Distance
...

Origin
...


A scale of measurement allows the investigator to make comparisons of amounts and changes in
the variable being measured
...

222
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Primary Types of Scales
To many people, the term scale suggests such devices as a bathroom scale, pan balances,
yard sticks, gasoline gauges, measuring cups, and similar instruments for finding length, weight,
volume, and the like
...
In the behavioral sciences
(including marketing research), however, we must frequently settle for less-precise data
...

Each scale possesses its own set of underlying assumptions about order, distance and
origin, and how well the numbers correspond with real-world entities
...
One example is the
measurement of color
...

The specification of scale is extremely important in all research, because the type of measurement
scale dictates the specific analytical (statistical) techniques that are most appropriate to use in
analyzing the obtained data
...
They support
only the most basic analyses
...
The nominal scale does not possess order, distance, or origin
...
We have a one-to-one correspondence between
number and player and are careful to make sure that no players receive the same number (or that
a single player is assigned two or more numbers)
...

It should be clear that nominal scales permit only rudimentary mathematical operations
...
The usual statistical operations involving the calculations of
means, standard deviations, etc
...

Ordinal Scales
Ordinal scales are ranking scales and possess only the characteristic of order
...
For example, a respondent may be asked to rank a group of floor polish brands
according to “cleaning ability”
...
Note, however, that the
mere ranking of brands does not quantify the differences separating brands with regard to
cleaning ability
...
In dealing
with ordinal scales, statistical description can employ positional measures such as the median,
quartile, and percentile, or other summary statistics that deal with order among
...
Also, like the nominal scale, arithmetic averaging is
not meaningful for ranked data
...
Smith and Gerald S
...
This type of scale possesses the
properties of order and distance, but the zero point of the scale is arbitrary
...
While an arbitrary zero
is assigned to each temperature scale, equal temperature differences are found by scaling equal
volumes of expansion in the liquid used in the thermometer
...

An example should make this point clearer
...
Converting from Fahrenheit to Centigrade, we find that the
corresponding temperatures on the centigrade scale are 10°C and –3
...
We can say, however, that differences between values on different temperature scales
are multiples of each other
...
The corresponding differences on the Centigrade scale are 10°C – (–17
...
7°C and –
3
...
7°C) = 13
...

Interval scales are unique up to a transformation of the form y = a + bx; b > 0
...

For example, we can convert from a Fahrenheit to Celsius using the formula:
TC = 5/9 (TF – 32)
Most ordinary statistical measures (such as arithmetic mean, standard deviation, and
correlation coefficient) require only interval scales for their computation
...
These are scales like length and weight that possess a unique zero point,
in addition to equal intervals
...

An example of ratio-scale properties is that 3 yards is three times 1 yard
...
It is easy to move from one scale to
another merely by applying an appropriate positive multiplicative constant; this is the practice
followed when changing from grams to pounds or from feet to inches
...

From a purely mathematical point of view, you can obviously do arithmetic with any set
of numbers—and any scale
...
As we select more powerful measurement scales, our abilities to predict, explain, and
otherwise understand respondent ratings also increase
...
Smith and Gerald S
...
1

Scale
Nominal
Ordinal

Scales of Measurement

Mathematical
Group Structure
Permutation group
y = f(x), where f(x) means any
one-to-one correspondence
Isotonic group
y = f(x), where f(x) means any
strictly increasing function

Interval

General linear group
y = a+bx
b>0

Ratio

Similarity group
y = cx
c>0

Permissible Statistics

Typical Elements

Mode
Contingency Coefficient

Numbering of football players
Assignment of type or model
numbers to classes
Hardness of minerals
Quality of leather, lumber,
wool, etc
...
678)

Basic Question and Answer Formats
Underlying every question is a basic reason for asking it
...
Constructing a
question that reflects this reason will result in a higher probability that the desired response will
be obtained
...
2 shows nine different types of questions (based on the nature of content),
the broad reason underlying asking each type of question, and some examples of each type
...
2

Basic Question Types

Type of Question

Goal of Question

Positioning of Question

Factual or
behavioral
Explanatory

To get information
...
, about an object or topic
...

To introduce a thought of your own
...


Alternative

To get a decision or agreement
...

To take action
...


Questions beginning with what, where, when,
why, who and how
...


Is baseball more or less exciting to watch on
TV than soccer?

225
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Based on this structure, and the information in Table 10
...

2
...

4
...
3
Measurement
Scale*

Standard Answer Formats Based on Task
Format Type

Description

N,O,I

Select 1/n—pick-1:

The respondent is given a list of n options and is required to choose one option only
...

The respondent is asked to select k1 options in Category 1 and k2 options in Category 2
...

The respondent picks k items and allocates them into L buckets, then items allocated to each
bucket are assigned ranks
...


N,O
O,I

Select k1/n and Rank
k2/k1—pick and rank:
Integer Rating:

O,I

Continuous Rating:

R

Constant Sum:

N,O

Yes/No:

I

Integer—integer-#:

I,R

Real—real-#:

C

Character:

This question type is similar to pick-k, but in addition to selecting k1 options from a list of n
options, the respondent is then asked to rank some fraction, k2/k1 of those selected
...

Only integer responses are accepted
...
2 on a scale of 0 to 10)
...
) and is asked to distribute a total
of p points across those attributes
...

The respondent is asked for a fact that can be expressed in integer number form
...
Example: Age
...
Example: Income
...

The respondent types in a string of characters as a response
...


Multiple Integer
This question type is identical to integer-scale except that multiple questions (classified as
Ratings:
“options”) can appear on a single screen
...

Multiple Real
This question type is identical to real-scale except that multiple questions (classified as
I,R
Number Ratings:
“options”) can appear on a single screen
...

*Legend: N = Nominal, O = Ordinal, I = Interval, R = Ratio, C = Alpha-Numeric Text Characters
I

Free Answer or Open-Ended Text Questions/Answers

The free answer (or open-ended text question) has no fixed alternatives to which the
answer must conform
...

Interviewers are usually instructed to make a verbatim record of the answer
...
As such, this form of question provides the opportunity for greater
ambiguity in interpreting answers
...
Smith and Gerald S
...
I never thought much about it
...
Bottles break, though
...
One
practitioner has gone so far as to suggest that questionnaires should clearly instruct interviewers
to probe only once for additional information, and to continue to probe for clarity until the
interviewer understands a respondent’s reply
...
1), we may tentatively conclude that
the free-answer question provides the lowest probability of the questions being ambiguous, but
the highest probability of the answers being ambiguous,
...
1 Open-Ended Questions and Answers
The advantages of the open-ended format are considerable, but so are its disadvantages
(Sudman and Bradburn, 1982)
...

It also allows respondents to make distinctions that are not usually possible with the fixed alternative
formats, and to express themselves in language that is comfortable for them and congenial to their
views
...

The richness of the material can be a disadvantage if there is need to summarize the data into
simple response categories
...

Open-ended questions also take somewhat more time and psychological work to answer than
closed questions
...
Open-ended response formats may work better
with telephone interviews, where a close supervision of interview quality can be maintained, although
there is a tendency for shorter answers to be given on the telephone
...


Dichotomous and Multiple-Choice Answers
The select k of n format is the workhorse of survey building, and provides the general
form for both dichotomous and multiple-choice answer types
...
Smith and Gerald S
...
A dichotomous question has two fixed answer alternatives of the type “Yes/No”, “In
favor/Not in favor”, “Use/Do not use”, and so on
...
Multiple-choice questions are
simply an extension of the dichotomous question that have more answer points and often take the
form of an ordered or interval measurement scale
...
For example, an agreement scale could have three, five, or
seven available answers
...

Select Exactly k of n Answers

When questions are developed that accept or require multiple responses within a set of
answers, the form “exactly k of n” or “as many as k of n” can be used
...
In this case,
the data collected would be of type categorical or even loosely ordered if presence or absence of
a characteristic is being measured (data is coded as 0 if not selected, and 1 if selected)
...
In these cases, the respondent is asked to select as many as k of
the n possible answers, where k can be any number from 2 to n
...
The question might be reworded to read something like
...

Whitlark and Smith (2004) show an application of pick k of n data that asks respondents
to pick a small number of attributes that they feel best describe a brand from a list of 10, 20, or
even 30 attributes
...
Smith and Gerald S
...
In an online survey environment, respondents can quickly
scan down columns or across screens and quickly complete the pick data task for a familiar
brand, thereby saving time and reducing respondent fatigue and dropout rates
...
Whitlark and Smith (2004) show that when
respondents are asked to pick from one third to one half of the viewed items, the pick k of n data
can be superior to scaled data in terms of reliability and power to discriminate between attributes
...
Whereas the categorical data associated with many dichotomous or multiple-choice
items does not permit us to say that one item is greater than another, rank-order data allows for
the analysis of differences
...
The first,
second, and so forth up to the nth item would be ordered
...
Typical
questions might include identifying preference rankings, or attribute associations from first to
last, most recent to least recent or relative position (most, next most, and so forth, until either a
set number of items is ordered or all items may be ordered)
...
Randomization of the answer list
order helps to control for presentation order bias
...

Tied rankings are another issue to be considered for rank-order questions
...
In general, this is not a good
idea because it weakens the data
...

Rank-order questions are generally a difficult type of question for respondents to answer,
especially if the number of items to be ranked goes beyond five or seven
...
This type of question is used when you are
relatively sure of the answer set (i
...
, reasons for purchase, or you want to evaluate a limited
number of reasons that you believe are important)
...
Smith and Gerald S
...
We focus on broad concepts of attitude scaling—the study of scaling for
the measurement of managerial and consumer or buyer perception, preference, and motivation
...
The stimuli may be alternative products
or services, advertising copy themes, package designs, brand names, sales presentations, and so
on
...

Scaling procedures can be classified in terms of the measurement properties of the final
scale (nominal, ordinal, interval, or ratio), the task that the subject is asked to perform, or in still
other ways, such as whether the scale measures the subject, the stimuli, or both (Torgerson,
1958)
...
) in terms of their mechanics and assumptions regarding their
scale properties
...

Techniques such as Thurstone Case V scaling, semantic differential, the Likert summated scale,
and the Thurstone differential scale are illustrated
...

230
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Advanced Ordinal Measurement Methods
The variety of ordinal measurement methods includes a number of techniques:
Paired comparisons
Ranking procedures
Ordered-category sorting
Rating techniques

We discuss each of these data collection procedures in turn
...
If, for example, six laundry detergent brands are
to be compared for “sudsiness”, a full set of paired comparisons would involve (n x n – 1)/2 = (6
5) / 2, or 15, paired comparisons (if order of presentation is not considered)
...

A sample question format for paired comparisons is shown in Table 10
...
The order of
presentation of the pairs and which item of a pair is shown first are typically determined and/or
presented randomly
...

Table 10
...
Smith and Gerald S
...


Ranking Procedures
Ranking procedures require the respondent to order stimuli with respect to some
designated property of interest
...

Similarly, ranking can be used to determine key attributes for services
...
One question required ranking:

A variety of ordering methods may be used to order k items from a full set of n items
...
The various ordering methods may pre-specify the value of k
(“order the top three out of six brands with respect to sudsiness”) as illustrated by the Subaru
study, or allow k to be chosen by the respondent (“select those of the six brands that seem to
exhibit the most sudsiness, and rank them”)
...
For example, if it is desired that a respondent rank all items in a longer list
of items, the pick-group-rank procedure may be used
...
Then, the request is made to rank the items within each pile
...

232
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Ordered-Category Sorting
Pick-Group-Rank is one of a variety of data collection procedures that have as their
purpose the assignment of a set of stimuli to a set of ordered categories
...
free versus forced assignment of names to grouping categories
2
...
the assumption of equal intervals between category boundaries versus the
weaker assumption of category boundaries that are merely ordered with
regard to the attribute of interest
In ordinal measurement methods one assumes only an ordering of category boundaries
...
Ordered-category sorting appears especially useful when the researcher is
dealing with a relatively large number of stimuli (over 15 or so) and it is believed that a subject’s
discrimination abilities do not justify a strict (no ties allowed) ranking of the stimulus objects
...
In some cases, the scaled responses are considered
by the researcher to be only ordinal, while in other cases, the researcher treats them as interval233
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

or ratio-scaled
...

The rating task typically involves having a respondent place that which is being rated (a
person, object, or concept) along a continuum or in one of an ordered set of categories
...
The task of rating is one of the most popular and easily applied data collection
methods, and is used in a variety of scaling approaches, such as the semantic differential and the
Likert summated scale
...
In monadic scaling, each object is
measured (rated) by itself, independently of any other objects being rated
...
For example, a
recent in-flight survey conducted by United Airlines asked the following questions:

The rating is monadic
...
But there is a risk of lack
of differentiation among the scores when the particular attributes are worded positively or are
positive constructs, such as values, and the respondents end-pile their ratings toward the positive
end of the scale
...

McCarty and Shrum (2000) offer an alternative to simple rating
...
The
remaining values were then rated
...

234
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Respondents may have trouble choosing a rating scale value on the high end, but not so at
the low end
...
Semon
(1999) suggests that one way to find the real difference in perception or attitude is to ask each
respondent three questions at the start of an interview:
1
...
How do you rate the best brand you know about?
3
...
A respondent’s actual ratings can be
translated into responses relative to one or more of these anchors to produce real-meaning
relative ratings that can be reliably aggregated and analyzed without depending upon
assumptions that may be questionable
...
Often two or more
of these formats appear together, as illustrated in Figures 10
...
2
...

Figure 10
...


b
...

e
...


In many instances where rating scales are used, the researcher assumes not only that the
items are capable of being ranked, but also that the descriptive levels of progress are in equalinterval steps psychologically
...
1 may be treated—sometimes erroneously—as interval- or ratio-scaled data
...

Treating rating scales as interval or ratio measurements is a practice that is well documented and
widespread
...
Research evidence supports this practice, in that often when ordinal
data are treated as interval and parametric analysis are used, the conclusions reached are the
same as when the data are treated as ordinal and tested using non-parametric analyses
...
Smith and Gerald S
...
2

A Rating Thermometer

Constructing a Behaviorally Anchored Rating Scale
One type of itemized rating scale that has merit in cases where leniency error (lack of
discrimination) may be troublesome is the behaviorally-anchored rating scale, or BARS (see
Figure 10
...
This scale uses behavioral incidents to define each position on the rating scale
rather than verbal, graphic, or numeric labels
...
Developing scales such as these requires a
great amount of testing and refinement to find the right anchors for the situation under
examination
...
3

Behaviorally Anchored Rating Scale (BARS)

The basic process of developing a behaviorally anchored rating scale consists of four steps:
236
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

1
...
Item generation—statements must be generated describing actual behaviors that would
illustrate specific levels of the construct for each dimension identified
3
...
Scale construction—lay out the scale with behavioral statements as anchors

In following this process, sets of judges are used
...
Thus, they should be reserved for those applied
settings where they can minimize the errors they are designed to curtail, especially leniency
error
...
A BARS was used for one critical measure of how well elderly
members of the household were able to perform everyday living activities:

Table 10
...

Table 10
...
Should negative numbers be used?
2
...
Related to the number of categories is: Should there be an odd number or an even number?
That is, should a neutral alternative be provided?
4
...
Is it desirable to not force a substantive response by giving an opportunity to indicate “don’t
know,” “no opinion,” or something similar?
6
...
How does one examine raters’ biases—for example, the tendency to use extreme values or,
perhaps, only the middle range of the response scale, or to overestimate the desirable
features of the things they like (i
...
, the generosity error)?
8
...
How anchoring phrases for the scale’s origin should be chosen?

Some research on these questions has been conducted showing that errors are not made
when a neutral option is provided
...
g
...
Smith and Gerald S
...
Expected voting in a
survey of voters is an example
...
When using importance scales for
attributes, the alternatives provided may be “very important”, “important”, “neither important
nor unimportant”, “unimportant”, and “very unimportant”, or additional categories may be
included
...
He argues that importance is not a bipolar concept
...
Although this appears to have conceptual
appeal, researchers continue to use successfully use importance scales from some mid-point—
specified or implied
...
Isolating any findings of unimportance
2
...
Accurately interpreting the relative nature of importance findings
Answers to questions such as these will vary by the researcher’s approach, and by the
problem being studied
...

Exhibit 10
...
Among the areas of most concern are better scaling techniques for
measuring children’s product preferences
...

Although these scales have done well in varied research applications, there is often a problem with
leniency that emerges, particularly when used with young children under the age of eight
...
If this is done for all products tested, the overall
sensitivity of existing (traditional) rating scales is lowered, resulting in inconclusive findings about
children’s preferences
...
The latter two scales are
238
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

taken up next
...

Interval/Ratio Procedures
Direct-judgment estimates, fractionation, constant sum and rating methods assume more
than ordinal properties about respondents’ judgments) are all variants of interval/ratio procedures
or metric measurement methods
...
In the case of continuous rating scales, the
respondent is free to choose his or her own number along some line that represents his or her
judgment about the magnitude of the stimulus relative to some reference points (Figure 10
...

These continuous scales work effectively in a semantic differential context (to be
discussed later in this chapter) (Albaum, Best, & Hawkins, 1981), and appear to be insensitive to
fluctuations in the length of the line used (Hubbard, Little, & Allen, 1989)
...
4, is nothing
more than a straight rating procedure, with the important addition that the ratings are now treated
as either interval- or ratio-scaled data (depending on the application)
...
4

Sample Interval-Ratio Scales

(a) Unlimited-Response Categories
Don’t
Like
At All

(c)

(b) Limited-Response Categories
Like
Very
Much

Fractionation

(d) Constant Sum

Result:
The Relative degree of sudsiness compared to Brand A
Brand
Response
B
0
...
80
D
2
...
50

Fractionation
Fractionation is a procedure in which the respondent is given two stimuli at a time (e
...
, a
standard laundry detergent and a test brand) and asked to give some numerical estimate of the
ratio between them, with respect to some attribute, such as sudsiness
...
After
239
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

this is done, a new test brand is compared with the same standard, and so on, until all test items
are judged
...
4 illustrates this procedure
...

For example, in an actual test of the attribute is sweetness of lemonade, the respondent may be
asked to add more sweetener until the test item is “twice as sweet” as the standard
...
In constant-sum methods the respondent is
given some number of points—typically 10 or 100—and asked to distribute them over some set
of stimuli or attribute alternatives in a way that reflects their relative importance or magnitude
(Figure 10
...
Constant sum forces the respondent to make comparative evaluations across the
stimuli, and effectively standardizes each scale across persons, since all scores must add to the
same constant
...

In summary, unlike ordinal measurement methods, the major assumption underlying
ratio/interval measurement methods) is that a unit of measurement can be constructed directly
from respondents’ estimates about scale values associated with a set of stimuli
...
The problems associated with Interval-Ratio Scaling methods include
the following:
1
...

2
...

3
...

4
...

These problems should not be treated lightly, but considered in the design of the question
and scale points
...
Moreover,
little additional work beyond averaging is required to obtain the unit of measurement directly
...
g
...
As such, a subjective ratio
scale is obtained
...
Smith and Gerald S
...

You find the results of a survey conducted just before you came on board
...
2, 1 vs
...
4, etc
...
But your manager specifically asked
that preference be displayed on a single continuous (interval or ratio scale)
...
6)…
...
One such
transformation, Thurstone’s Case V method, is capable of transforming ordinal data obtained
from ranking methods
...
However, in this case the model may be no more elaborate than averaging the
raw data across respondents and/or response occasions
...
We consider each of these techniques in turn
...
This model can also be used to scale
ranked data or ordered-category sorts
...
We shall first describe the general case and then concentrate on Case V, a special
version particularly amenable to application in marketing situations
...
Scale values may be estimated from data in which one individual makes many
repeated judgments on each pair of a set of stimuli or from data obtained from a group of
individuals with few or no replications per person
...
Assume that the survey
you found had asked 100 homemakers to compare five brands of “fortified juice” with respect to
“overall preference of flavor”
...
Table 10
...

From this table we see that 69 percent of the respondents preferred Juice C to Juice A and
the remainder, 31 percent preferred Juice A to Juice C (if we arbitrarily let column dominate
row)
...
7) to 0
...
From the data of this table we next
241
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

prepare Table 10
...
These Zvalues were obtained from Table A
...
If the proportion is
less than 0
...
5, the Z-value
carries a positive sign
...
The Thurstonian model assumes normally distributed scale
differences in mean = 0 and standard deviation = 1
...

For example, from Table 10
...
82
...
This value (labeled Z
in the standard unit normal table of Table Appendix A
...
92
...
92
...
6
are obtained in a similar manner, a minus sign being prefixed to the Z-value when the proportion
is less than 0
...

Column totals are next found for the entries in Table 10
...
Scale values are obtained from
the column sums by taking a simple average of each column’s Z-values
...
7, we note that the sum of the Zs for the first column (Juice A) is –0
...
The average Z
for column A is simply:
0
...
072
5
This scale value expresses Juice A as a deviation from the mean of all five scale values
...
Similarly, we find the average Z-value for each of the remaining four columns of
Table 10
...

Table 10
...
7
Brand
A
B
C
D
E
Total
Mean (Z)
R

Observed Proportions Preferring Brand X to Brand Y
Proportion Preferring Column Brand to Row Brand

A
0
...
18
0
...
75
0
...
82
0
...
73
0
...
85

C
0
...
27
0
...
84
0
...
25
0
...
16
0
...
41

E
0
...
15
0
...
59
0
...
6
Brand
A
B
C
D
E
0
0
...
50
-0
...
39
-0
...
61
-1
...
04
-0
...
61
0
-0
...
67
0
...
48
0
...
23
0
...
04
0
...
23
0
-0
...
05
1
...
37
-1
...
072
0
...
31
-0
...
374
0
...
484
0
...
3

Next, since the zero point of an interval scale is arbitrary, we can transform the minimum
scale so that it becomes zero
...
674), we force it
to become the reference point (or origin) of zero by adding
...
We then simply add 0
...
These are
denoted by R* and appear in the last row of Table 10
...

The scale values of Juices A through E indicate the preference ordering B > C > A > E >
D
...
3 times the difference in “goodness of
242
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

flavor” between Juices C and A, since
B – A = 2
...
484 – 0
...
3(0
...
602)
0
...
3(0
...

The test of this model is how well scale values can be used to work backward—that is, to
predict the original proportions
...
For any specific brand, the highest mean absolute proportion discrepancy is 0
...
Moreover, the overall mean absolute discrepancy is only
...
Even the simplest
version (Case V) of the Thurstonian model leads to fairly accurate predictions
...

The Semantic Differential

The semantic differential (Osgood, Suci, & Tannenbaum, 1957) is a ratings procedure
that results in (assumed interval) scales that are often further analyzed by such techniques as
factor analysis (see Chapter 14)
...
It is simply assumed that the raw data are
interval-scaled; the intent of the semantic differential is to obtain these raw data for later
processing by various multivariate models
...
e
...

As shown in Figure 10
...
For each
pair of bi-polar adjectives (e
...
, powerful/weak), the respondent is asked to judge the concept
along the seven-point scale with implicit descriptive phrases
...
5

Corporate Profile Obtained by the Semantic Differential

In practice, however, profiles would be built up for a large sample of respondents, with
many more bipolar adjectives being used than given here
...
6, the responses can be quantified under the
assumption of equal-appearing intervals
...
For example, Figure 10
...
The average score for the respondents show that
the Company X is perceived as very weak, unreliable, old-fashioned, and careless, but rather
warm
...


243
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 10
...
These scales
are developed for particular context areas, so the scales have more meaning to respondents, thus
leading usually to a high degree of reliability
...
7 apply to the semantic
differential
...
As Figure 10
...

Figure 10
...
Smith and Gerald S
...
Comparative studies of data quality (including reliability) seem to indicate that the
choice of a format may be appropriately made on the basis of ease of subject understanding, ease
of coding and interpretation for the researcher, ease of production and display, and cost
...

A Concluding Remark

The semantic differential technique is appropriate for use in a variety of applications:
Comparing corporate images, both among suppliers of particular products and against an ideal
image of what respondents think a company should be
Comparing brands and services of competing suppliers
Determining the attitudinal characteristics of purchasers of particular product classes or brands
within a product class, including perceptions of the country of origin for imported products
Analyzing the effectiveness of advertising and other promotional stimuli toward changing
attitudes

The comparatively widespread use of the semantic differential by marketing researchers
suggests that this method provides a convenient and reasonably reliable way for scaling stimuli
(scaling images of brands, corporations, services, etc
...

Techniques for Scaling Respondents
In contrast to the approaches for scaling stimuli just discussed, researchers also have available
techniques whose primary purpose is to scale respondents along some attitude continuum of
interest
...
Each of these is described in turn
...
To illustrate, assume that the researcher wishes to
scale some characteristic, such as the public’s attitude toward travel and vacations
...
8
...
Smith and Gerald S
...
Each
subject would be asked to indicate their agreement with the statement
...
Since, we reverse scaled items 2, 4, 5, and 6 on “unfavorable” statements, we would
reverse the order of the scale values so as to maintain a consistent direction ( +2 would stand for
“strongly disagree,” and so on)
...
This person’s score would be:
– 2 + 0 – 1 – 2 – 2 – 2 + 0 = –9
This listing indicates that the second respondent would be ranked “lower” than the first—
that is, as having a less-favorable attitude regarding travel and vacations
...

Figure 10
...
3 are
typically carried out
...
3 Steps in Constructing a Likert Summated Scale
1
...
g
...

2
...
No attempt is made to scale the items; however, a pretest is
conducted that involves the full set of statements and a limited sample of respondents
...

3
...
Smith and Gerald S
...
Each response is given a numerical weight (e
...
, +2, +1, 0,
+1 to +5)
...
The individual’s total-attitude score is represented by the algebraic summation of weights
associated with the items checked
...
For example, if a + 2
were assigned to “strongly approve/agree” for favorable items, a + 2 should be assigned to
“strongly disapprove/disagree” for unfavorable items
...
On the basis of the results of the pretest, the analyst selects only those items that appear to
discriminate well between high and low total scorers
...
Then, the mean differences
on each specific item are compared between these high and low groups (excluding the middle
50 percent of subjects)
...
The 20 to 25 items finally selected are those that have discriminated “best” (i
...
, exhibited the
greatest differences in mean values) between high versus low total scorers in the pretest
...
Steps 3 through 5 are then repeated in the main study
...
Nonetheless, two respondents could have the
same total score even though their response patterns to individual items were quite different
...
Moreover, the total score is sensitive to how the respondent reacts to the
descriptive intensity scale
...
e
...

Having positively and negatively worded statements hopefully forces respondents with strong
positive or negative attitudes to read carefully and use both ends of a scale
...
That is, a
“strongly agree” response to a positive statement and a “strongly disagree” to a negative
statement should be scored the same, and so forth
...
As an illustration, a non-directional format for one
item would be:
“How much pleasure do you get from traveling? [Very little…A great deal]”
In contrast, the normal Likert format for this item is:
“Traveling gives me a lot of pleasure [strongly agree, agree, neither agree nor
disagree, disagree, strongly disagree]”
Some final comments are in order
...
Because two persons with decidedly different
attitudes may agree on fact, it is recognized that direction is the only meaningful measure
247
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

obtained when using statements of facts
...
As stated earlier, this may lead to reluctance on the part
of respondents to either give extreme scores or use the extreme position on an individual scale
item (central tendency error)
...

The Q-Sort Technique
The Q-sort technique has aspects in common with the summated scale
...

In illustrating the Q-sort technique, assume that four respondents evaluate the test items
dealing with travel and vacations
...

The respondents are asked to sort items into:
Group 1
LEAST AGREED WITH
-1

Group 2
NEUTRAL ABOUT
0

Group 3
MOST AGREED WITH
+1

Suppose that the responses toward seven items by the four respondents, A, B, C, and D,
result in the following scale values:
Item
1
2
3
4
5
6
7

A
+1
0
+1
-1
0
–1
0

Respondent
B
+1
0
0
–1
0
–1
+1

C
–1
0
0
+1
0
+1
–1

D
–1
0
–1
+1
0
+1
–1

As can be noted, the respondent pairs A & B and C & D seem to be the “most alike” of
the six distinct pairs that could be considered
...
Typically, these additional steps are undertaken in Q-sort studies
...
A multi-item scale consists of a number of
closely related individual rating scales whose responses are combined into a single index,
composite score, or value (Peterson, 2000)
...
Multi-item scales are used when measuring complex psychological constructs that are not
easily defined by just one rating scale or captured by just one question
...
The first, and perhaps most critical,
step is to clearly and precisely define the construct of interest
...
Smith and Gerald S
...
This is followed by design and evaluation of
the scale
...

Along the way, a pilot study is conducted to further refine the scale and move toward the final
version
...
Of concern is construct
validation, in which an assessment is made that the scale measures what it is supposed to
measure
...

Norms describe the distributional characteristics of a given population on the scale
...
9)
...
Reliability is assessed by the scale’s
stability (test-retest reliability) and internal consistency reliability (coefficient alpha)
...

The scale must be appropriate to the population of people who use it, such as having an
appropriate reading level
...

Consider possible biasing factors and sensitive items
...
8 gives an example of a multi-item scale developed to measure consumer
ethnocentrism within a nation, the CETSCALE (Shimp & Sharma, 1987)
...
A Cetscale score for an individual
respondent is obtained as a sum of item ratings, and ranges from 17 to 119 with higher numbers
indicating greater consumer ethnocentrism
...

Table 10
...
American people should always buy American-made products instead of imports
...
Only those products that are unavailable in the United States should be imported
...
Buy American-made products
...

4
...

5
...

6
...

7
...

8
...

9
...

10
...

11
...

12
...

13
...

14
...

15
...

16
...

17
...

NOTE: Items composing the 10-item reduced version are items 2, 4 through 8, 11, 13, 16, and 17
...
Smith and Gerald S
...
We still do
not know, in many cases, how to effectively translate verbalized product ratings, attitudes about
corporations, and so on into the behavioral and financial measures required to evaluate the
effectiveness of alternative marketing actions
...
Consider that the following question differing only in the use of the words
“should”, “could” and “might” was shown to three matched samples of respondents (Payne,
1951, pp
...

Do you think anything should be done to make it easier for people to pay doctor or
hospital bills? (82 percent replied “Yes”
...
These three words are sometimes used as synonyms, and
yet at the extreme, responses are 19 percentage points apart
...
1 percent of respondents replied “too little”, but when the label was
changed to assistance to the poor, 62
...
Questions portraying a more
descriptive and positive position may show a large difference in the evaluation score
...
In
this section we discuss issues of question structure and form that can greatly influence and
improve the quality of your questionnaire
...
Writing questions is an
art, which like all arts requires a great amount of work, practice, and help from others
...
4 we provide an overview of the common pitfalls we often see in "bad questionnaires" that
lead to various forms of ambiguity
...
Smith and Gerald S
...
4

Key Considerations for Reducing Question Ambiguity

1
...
One study “should”, “could” and “might” was shown
to three matched samples of respondents (Payne, 1951, pp
...
Do you think anything should
be done to make it easier for people to pay doctor or hospital bills? (82 percent replied “Yes”
...
These three words are sometimes used as synonyms, and yet
at the extreme, responses are 19 percentage points apart
...

2
...
Slight wording changes can produce great
differences in results
...
should
...
have forced
the breakup of Microsoft Corporation)
...
Framing effects
...
Levin, Schneider, and Gaeth (1998) and Levin et al
...

Goal framing effects occur when a persuasive message has different appeal depending on
whether it stresses the positive consequences of performing an act to achieve a particular
goal or the negative consequences of not performing the act
...

Which type of potential framing effects should be of concern to the research designer depends
upon the nature of the information being sought in a questionnaire
...
” This is an example of attribute framing
...
More detailed discussion of these effects is given by
Hogarth (1982)
...
Misplaced questions
...
In
general, a funnel approach is advised
...
Then more
specific questions, followed by more general easy to answer questions (like demographics) at
the end of the questionnaire
...
Mutually non-exclusive response categories
...
Non-exclusive answers frustrate the
respondent and make interpretation difficult at best
...
Nonspecific questions
...
do I like what? Taste,
texture, nutritional content, Vitamin C, the current price, concentrate, fresh squeezed? Be
specific in what you want to know about
...

7
...
Asking about caloric content, acrylamide, phytosterols, and
other industry specific jargon and acronyms are confusing
...


251
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

8
...

What suggestions do you have for improving tomato juice? The question is about taste, but the
respondent may offer suggestions about texture, the type of can or bottle, mixing juices, or
something related to use as a mixer or in recipes
...
Forcing answers
...
Privacy is an important issue to most people
...

10
...
Do you have all of the options covered? If you are unsure, conduct a
pretest using the "Other (please specify) __________" option
...

11
...
Unbalanced scales may be appropriate for some situations and biased in
others
...
However, we expect all hospitals to offer
good care and may use a scale of excellent, very good, good, fair
...

12
...
What is the fastest and most convenient Internet service for you?
The fastest is certainly not the most economical
...

13
...
Make sure answers are independent
...

14
...
Multiple choice questions are the longest and most complex
...
When you Increase the length of questions and surveys,
you decrease the chance of receiving a completed response
...
Questions on future intentions
...
Predictions
are rarely accurate more than a few weeks or in some case months ahead
...
However the measuring instrument does not include all of
the possible items that could have been included
...
Do the scales really measure what we are trying to measure?
2
...
If we have a variety of scaling procedures, are respondents consistent in their scoring
over those scales that purport to be measuring the same thing?

By solving these problems, we establish the validity and reliability of scaling techniques
...
There is little documented research on issues of
measure reliability and validity for time-series analysis
...
Smith and Gerald S
...
The
data must be unbiased and relevant to the characteristic being measured
...
Systematic error may arise from
the instrument itself, the user of the instrument, the subject, or the environment in which the
scaling procedure is being administered
...

The validity of a measuring instrument hinges on the availability of an external criterion
that is thought to be correct
...

What makes the problem even more difficult is that the researcher often is not interested in the
scales themselves, but the underlying theoretical construct that the scale purports to measure
...

In testing the validity of a scale, the researcher must be aware that many forms of validity
exist, including (1) Content validity, (2) Criterion validity, and (3) Construct validity
...
It is essentially judgmental and is ordinarily measured
by the personal judgments of experts in the field
...
Closely related to this approach for assessing content validation is a method
involving known groups
...
If the scale does not discriminate
between the two groups, then its validity with respect to measuring brand attitude is highly
questionable
...

Face Validity is a preliminary or exploratory form of content validity
...
A simple approach is to show the measurement instrument to a
convenient group of untrained people and ask whether or not the items look okay (Litwin, 2003)
...
This type
of validation is derived from the careful definition of the continuum of a scale and the selection
of items to be scaled
...
Because things often do not
turn out to be as obvious as believed, it is wise for the marketing researcher not to rely on logical
validation alone
...
Since the product represented a major change in taste, thousands of
consumers were asked to taste New Coke
...
With such a favorable reaction, why did the decision to introduce the product turn out to
be a mistake? Executives of the company acknowledge that the consumer survey conducted
omitted a crucial question
...
In short, they were not asked if they would buy the
new product in place of the old one
...
Smith and Gerald S
...
They question if the instrument works and are better
decisions can be made with it than without it?
The New Coke example also illustrates a case of poor predictive validity
...

In concurrent validity, a secondary criterion, such as another scale, is used to compare
results
...
Often product
researchers will ask a question like “Overall, how much do you prefer Brand A soft drink?”, and
then follow with another question such as, “Given the following four brands, indicate the
percentage of your total soft drink purchases that you would make for each brand
...

Construct Validation

In construct validation the researcher is interested both in the question, “Does it work?”
(i
...
, predict), and in developing criteria that permit answering theoretical questions of why it
works and what deductions can be made concerning the theory underlying the instrument
...

Convergent Validity: The correspondence in results between attempts to measure the
same construct by two or more independent methods
...

Discriminant Validity: Refers to properties of scaling procedures that do differ when they
are supposed to—that is, in cases where they measure different characteristics of stimuli and/or
subjects
...
Discriminant validity concerns the extent
to which a measure is unique (and not simply a reflection of other variables), and as such it
provides the primary test for the presence of method variance
...
In nomological validity
the researcher attempts to relate measurements to a theoretical model that leads to further
deductions, interpretations, and tests, gradually building toward a nomological net, in which
several constructs are systematically interrelated
...
Specifically, more
emphasis should be placed on the theories, the processes used to develop the measures, and the
judgments of content validity
...
A scale may be reliable but not valid
...
An unreliable scale cannot be a valid one
...
1099) states that “according to classical test theory, highly reliable measures
are necessary, but not sufficient, for demonstrating high construct validity or high criterion
validity
...
Smith and Gerald S
...
Just how reliable a scaling procedure turns
out to be will depend on the dispersion of the characteristic in the population, the length of the
testing procedure, and its internal consistency
...
They further concluded that sampling
characteristics and measurement development processes had little impact
...
The
basics of reliability in a marketing context are reviewed by Peter (1979)
...
Do we achieve consistent results, assuming that the relevant characteristics of
the subjects are stable over trials? One potential problem, of course, is that the first measurement
may have an effect on the second one
...
If at all possible, the researcher should allow a minimum of
two weeks to elapse between measurements
...

Alternative Forms

The alternative forms method attempts to overcome the shortcomings of the test-retest
method by successively administering equivalent forms of the measure to the same sample
...
The forms of the
measurement device may be given one after the other or after a specified time interval,
depending upon the investigator’s interest in stability over time
...

Internal Consistency

Internal consistency refers to estimates of reliability within single testing occasions
...
The basic form of this method is split-half reliability, in which
items are divided into equivalent groups (say, odd- versus even-numbered questions, or even a
random split) and the item responses are correlated
...

A potential problem arises for split-half in that results may vary depending on how the
items are split in half
...
Whenever possible, alpha should be used as a measure of the
internal consistency of multi-item scales
...
One caution
however, is that there should be a sufficient number of items in the measure so that alpha
becomes meaningful
...
Although there is no generally acceptable heuristic
covering the number of items, common sense would indicate that the minimum number of items
should be four or perhaps even six
...
Smith and Gerald S
...
e
...
Consequently, when interpreting an obtained alpha, the number of items must
always be kept in mind
...
70
...

In many projects, measurements or evaluations are made by more than a single evaluator
...
In these situations, the
researcher is interested in the reliability of these evaluations
...
The most common measure used is a correlation (Litwin, 2003)
...
One question that has not been answered is,
“What is a satisfactory level of reliability, or what minimum level is acceptable?” There is no
simple definitive answer to this question
...
In
trying to arrive at what constitutes satisfactory reliability, the investigator must at all times
remember that reliability can affect certain qualities of a study including (1) Validity, (2) The
ability to show relationships between variables, and (3) The making of precise distinctions
among individuals and groups
...
We discussed the role of
definitions and defined concepts, constructs, variables, operational definitions, and propositions
...
Also discussed, but rather briefly, were alternative sources that cause
variations within a set of measurements derived from a single instrument
...
Advanced
scaling techniques for scaling stimuli and respondents were also discussed
...


256
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

REFERENCES
Albaum, G
...
, & Hawkins, D
...
Continuous vs
...

Psychological Reports, 49, 83–86
...
O
...
G
...
Handbook of Marketing Scales (2nd ed
...

Churchill, G
...
, Jr
...
P
...
Research design effects on the reliability of
rating scales: A meta analysis
...

Coombs, C
...
(1964)
...
New York: Wiley
...
(1993)
...
Journal of
Applied Psychology, 78(1), 98–104
...
J
...
Coefficient alpha and the internal structure of tests
...

Davis, D
...
(1997)
...
Public Opinion Quarterly, 61, 183–207
...
, Brockett, P
...
, & Zatarain, J
...
The golden numerical comparative scale
format for economical multi-object/multiattribute comparison questionnaires
...

Guttman, L
...
Measuring the true-state of opinion
...
Ferber & H
...
), Motivation and
market behavior
...
Irwin, 393–415
...
, & Case, P
...
Testing thirteen attitude scales for agreement and brand determination
...

Hogarth, R
...
(Ed
...
Question framing and response consistency
...

Hubbard, R
...
L
...
J
...
Are responses measured with graphic rating scales
subject to perceptual distortion? Psychological Reports, 69, 1203–1207
...
, & Duhachek, A
...
Applying confidence intervals to coefficient alpha
...

Kerlinger, F
...
Foundation of behavioral research (2nd ed
...

Levin, I
...
, Gaeth, G
...
, Albaum, G
...
(2001)
...
Asia Pacific Journal of
Marketing and Logistics, 13(2), 64–71
...
P
...
L
...
J
...
All frames are not created equal: A
typology of framing effects
...

Likert, R
...
The method of constructing an attitude scale
...
Fishbein (Ed
...
New York: Wiley, pp
...


257
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Litwin, M
...
(2003)
...
Thousand Oaks, CA:
Sage
...
A
...
J
...
The measurement of personal values in survey research: A test of
alternative rating procedures
...

Morgan, M
...
Be careful with survey data
...

Osgood, C
...
, Suci, G
...
, & Tannenbaum, P
...
(1957)
...
Urbana:
University of Illinois Press
...
P
...
A
...
(1986, February)
...
Journal of Marketing Research, 23, 1–10
...
A
...
Creating effective questionnaires
...

Payne, S
...
(1951)
...
Princeton, NJ: Princeton University Press
...
A
...
The effect of question wording on public support for government spending
...

Reuman, D
...
(1982)
...
Journal of Personality and Social Psychology, 43(5), 1098–
1110
...
T
...
Scale ratings always betrayed by arithmetic
...

Semon, T
...
(2001, October 8)
...
Marketing News, 35, 9
...
A
...
(1987, August)
...
Journal of Marketing Research, 24, 280–289
...
E
...
Summated rating scale construction: An introduction
...

Stevens, S
...
(1946)
...
Science, 103, 677–680
...
, & Bradburn, N
...
Asking questions
...

Thurstone, L
...
The measurement of values
...

Torgerson, W
...
(1958)
...
New York: Wiley
...
E
...
E, Jr
...
(2000)
...
Marketing Letters, 11(2), 177—191
...
, & Smith, S
...
"Pick and Choose “ Marketing Research Volume 16, Issue 4, Pages 814, American Marketing Association, Chicago, December, 2004
Wong, N
...
, & Burroughs, J
...
(2003, June)
...
Journal of Consumer Research, 30, 72–
91
...
Smith and Gerald S
...
For behavioral researchers, this scientific inquiry translates into a desire
to ask questions about the nature of relationships that affect behavior within markets
...

The first stage in the analysis process is identified to include editing, coding, and making
initial counts of responses (tabulation and cross tabulation)
...

In formulating hypotheses the researcher uses “interesting” variables, and considers their
relationships to each other, to find suggestions for working hypotheses that may or may not have
been originally considered
...
A
parameter is a summarizing property of a collectivity—such as a population—when that
collectivity is not considered to be a sample (Mohr, 1990, p
...

Although the sequence of procedures, (a) formulating hypotheses, (b) making inferences
and (c) estimating parameters is logical, in practice these steps tend to merge and do not always
follow in order
...
Similarly, not
all of the steps are always required in a particular project; the study may be exploratory in nature,
which means that it is designed more to formulate the hypotheses to be examined in a more
extensive project, than to make inferences or estimate parameters
...
Tabulation: identifying appropriate categories for the information desired, sorting the data by
categories, making the initial counts of responses, and using summarizing measures to provide
economy of description and thereby facilitate understanding
...
Formulating additional hypotheses: using the inductions derived from the data concerning the
relevant variables, their parameters, their differences, and their relationships to suggest working
hypotheses not originally considered
...
Making inferences: reaching conclusions about the variables that are important, their parameters,
their differences, and the relationships among them
...
Smith and Gerald S
...

2
...

4
...
Define appropriate categories for coding the information collected
...
Assign codes to the respondent’s answers
...
Enter the data into the computer and create a data file
...
Check the data file for errors by performing a
simple tabulation analysis to identify errors in coding or data entry
...

5
...
New variables may be computed by data manipulations that multiply,
sum, or otherwise transform variables
...
Weight Data Subclasses
...

7
...
Summarize the responses to each variable included in the analysis
...
A more in
depth discussion of survey-based data management is provided by Fink (2003, Chap
...

Defining Categories

The raw input to most data analyses consists of the basic data matrix, as shown in Table
11
...
In most data matrices, each row contains a respondent’s data and the columns identify the
variables or data fields collected for the respondent
...
This analysis is often done simply because we want to summarize the
meaning of the entire column of values
...
Because we summarize the data and
make inferences from it, it is doubly important that the data be accurate
...
The identification of response categories early in the study has several
advantages
...
It often leads to improvements in the questionnaire or observation forms
...
Editing problems are also reduced
...
Depending on the data collection method, data code sheets can be prepared and precoded
...
The major data analysis software programs read data files and then display
them in a spreadsheet-like database (see Table 11
...
Often the data are entered directly into a
Microsoft Excel spreadsheet for import into the statistical program to be used for analysis
...
Most of
today’s computer-based software for telephone (CATI) or internet surveys (Qualtrics
...
They not only define the question categories in the database but also
automatically build the database and record the completed responses as they are submitted
...
Smith and Gerald S
...
Response categories are
coded from 1 for the first category to the highest value for the last category
...

As desirable as the early definition of categories is, it can sometimes only be done after
the data have been collected
...

The selection of categories is controlled by both the purposes of the study and the nature
of the responses
...
Similarity of response within the category
...

2
...
Differences in category descriptions should be
great enough to disclose any important distinctions in the characteristic being examined
...
Mutually exclusive categories
...

4
...
The classification schema should provide categories for all
responses
...
Open-end questions, of
course, have their place in marketing research
...
All of this is by way of saying that any open-end question
should be carefully checked to see if a closed-end question (i
...
, check the appropriate box) can
be substituted without doing violence to the intent of the question
...

Editing and Coding
Editing is the process of reviewing the data to ensure maximum accuracy and clarity
...
Careful editing during the pre-test process will often catch misunderstandings of
instructions, errors in recording, and other problems so as to eliminate them for the later stages of
the study
...
Obviously, this has limited
application for printed questionnaires, though online or CATI surveys can be edited even when
data is being collected
...
If the sample is not large, a single editor usually edits all the data to reduce variation in
interpretation
...
In this way the same editor edits the same items on all forms, an arrangement that tends
to improve both consistency and productivity
...
Each collection form should be edited to ensure that data quality requirements are
fulfilled
...
Smith and Gerald S
...
Legibility of entries
...
Where not legible,
although it may be possible to infer the response from other data collected, where any real doubt
exists about the meaning of data it should not be used
...
Completeness of entries
...
It may mean either that the respondent could not or would not provide the answer,
that the interviewer failed to ask the question, or that there was a failure to record collected data
...
Consistency of entries
...
(If a
respondent family is indicated as being a non-watcher of game shows, for example, and a later
entry indicates that they watched Wheel of Fortune twice during the past week, an obvious
question arises as to which is correct
...
When discrepancies cannot be resolved,
discarding both entries is usually the wisest course of action
...
Accuracy of entries
...
Of particular importance is the detection of any repetitive response patterns in the reports of
individual interviews
...

Coding is the process of assigning respondent answers to data categories and numbers are
assigned to identify them with the categories
...
Sometimes these codes are printed on structured questionnaires and observation
forms before the data are collected
...

Post-coding is the assignment of codes to responses after the data are collected, and is
most often required when responses are reported in an unstructured format (open-ended text or
numeric input)
...

When not using CATI or online data collection technologies, a formal coding manual or
codebook is often created and made available to those who will be entering or analyzing the data
...
1 as
an illustration
...
The
editor-coder should be provided with written instructions, including examples
...
The coder also should be
aware of the computer routines that are expected to be applied, insofar as they may require
certain kinds of data formats
...
By comparing the results of the various coders, a process known as
determining inter-coder reliability, any inconsistencies can be brought out
...
Coding is an activity that should not be taken
262
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

lightly
...

Qualtrics has an interesting feature that uses a “wizard” to take a survey by selecting
random choices and following the various logic paths available
...

Figure 11
...
The tabulation of responses will
invariably reveal codes that are out of range or otherwise invalid
...
Some errors, such as the preceding one, represent entry of values that are
out-of-range or wild codes (Lewis-Beck, 1995, p
...
That is, the value is not one that has been
263
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

assigned to a possible response to a question
...
That is, in the coding shown in Figure 11
...
Hopefully, not too
many errors of this type occur
...
1
Object

1
2
3

...

i

...

n

Illustration of Data Matrix
1

2

3


...



...



...



...



...


Variable
…
…
…
…


...


j

…
…
…
…


...

…


...

…


...

…

m


...

…

An aspect of cleaning the data is dealing with missing data
...
One way of handling this is to use statistical
imputation
...
Researchers are mixed in their views about this
process
...
Statistical software programs can handle this either question-byquestion or by deleting the respondent with a missing value from all analyses
...
Yet another way is to simply assign the group’s mean value to the missing items
...

Another issue that can arise is how to deal with outliers (Fink, 2003, pp
...
Outliers
are respondents whose answers appear to be inconsistent with the rest of the data set
...

Regression analysis also can be used to detect outliers
...
Outliers
can be discarded from the analysis, but one must be careful to not throw out important and useful
information as would be the case when the outliers belong to a unique category of respondents
heretofore unidentified
...

Short of having two or more coders create the data file independently of each other and
then assessing intercoder reliability, there is not much that can be done to prevent coder error
except to impress upon coders the necessity of accurate data entry
...
Each error that is identified should
be traced back to the questionnaire to determine the proper code
...


264
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Tabulation: Basic Analysis

Tabulation may be thought of as the final step in the data collection process and the first
step in the analytical process
...

The most basic is the simple tabulation, often called the marginal tabulation and familiar
to all students of elementary statistics as the frequency distribution
...
A cross-tabulation involves the simultaneous counting of the
number of observations that occur in each of the data categories of two or more variables
...
2
...
A cross-tabulation is one of the more commonly employed and useful forms of
tabulation for analytical purposes
...
There is a common tendency for the researcher to decide that,
because cross-tabulations (and correlations) are so easily obtained, large numbers of tabulations
should be run
...
For 50 variables, for example, there are 1,225 different twovariable cross-tabulations that can be made
...

Table 11
...
From a practical
standpoint, hypotheses may be developed to solve a problem, answer a question, or imply a
possible course of action
...

Hypotheses must be empirically testable
...
The hypothesis may be
stated informally as a research question, or more formally as an alternative hypothesis, or in a
testable form known as a null hypothesis
...
75-84)
...
Research questions are not empirically testable,
but aid in the important task of directing and focusing the research effort
...
Smith and Gerald S
...
1 Development of a Research Question for Mingles
Mingles is an exclusive restaurant specializing in seafood prepared with a light Italian flair
...
, the owner and manager, has attempted to create an airy contemporary atmosphere that is
conducive to conversation and dining enjoyment
...

Barbara wants to track customer satisfaction with the Mingles concept, the quality of the
service, and the value of the food for the price paid
...
The
questionnaire asks, among other things:
“How would you rate the value of Mingles food for the price paid”?
The response format for the five answers appeared as:

When tabulated, the average response was found to be +0
...
43
...

Additional questions measure customer satisfaction by addressing “How satisfied Mingles customers
are with the concept, service, food, and value”
...
The logic of the null hypothesis is that if we
hypothesize no difference, and we “reject” the hypotheses if a difference is found
...
We may
only accept the null on a "tentative" basis because another testing of the null hypothesis using a
new sample may reveal that sampling error was present and that the null hypothesis should be
rejected
...
In this example, the difference between the population average, which is assumed to
be the middle scale value of 0 = "about average" and the sample's mean evaluation of Mingles
can be tested using the z distribution
...
Two examples:
1: Comparing two sample groups: H0: There is no difference in the value of the food for the
price paid as perceived by first time patrons and repeat patrons
...

2: Predicting intention to return to Mingles: H0: The perceived quality of service is not related
to the likelihood of returning to Mingles
...


Alternative hypotheses may be considered to be the opposite of the null hypotheses
...
Smith and Gerald S
...
Because population differences may exist, even if not verified by the
current sample data, the alternative form is considered to be empirically non-testable
...
3
...
3
Research
Question

Hypotheses and Research Questions
Purpose
Example
Express the purpose of the research
What is the perception
of Mingles customers
regarding the pricevalue of the food?

Decision
None used

Alternative
Hypothesis

The alternative hypothesis states the
specific nature of the hypothesized
relationship
...
e
...
The alternative hypothesis is
the opposite of the null hypothesis
...
(You can never reject
an alternative hypothesis unless you test
the population on all possible samples
...


Not tested because we
cannot reject
...


Null
Hypothesis

The null hypothesis is testable in the
sense that the hypothesized lack of
relationship can be tested
...
The Null hypothesis states
that there is no difference between
groups (with respect to some variable)
or that a given variable does not predict
or otherwise explain an observed
phenomena, effect or trend
...


We may reject a null
hypothesis (Find a
relationship)
...


The objectives and hypotheses of the study should be stated as clearly as possible and
agreed upon at the outset
...
However, a project will usually turn up new
hypotheses, regardless of the rigor with which it was planned and developed
...

In Chapter 2 it was pointed out that when the scientific method is strictly followed,
hypothesis formulation must precede the collection of data
...

New data must be collected prior to testing a new hypothesis
...
It is both acceptable and desirable to expand the analysis to
examine new hypotheses to the extent that the data permit
...
Smith and Gerald S
...
At the other extreme, a hypothesis may be supported by both the specific
variables tested and by other relationships that give similar interpretation
...
Between these extremes of nonsupport-support are outcomes of
indeterminacy: the new hypothesis is neither supported nor rejected by the data
...

In a position yet more extreme from scientific method, Selvin and Stuart (1966)
convincingly argue that in survey research, it is rarely possible to formulate precise hypotheses
independently of the data
...
Rather than having a single pre-designated hypothesis in mind, the analyst often works
with many diffuse variables that provide a slightly different approach and perspective on the
situation and problem
...
However, researchers must
resist the syndrome of “just one more question”
...

In a typical survey project, the analyst may alternate between searching the data
(analyzing) and formulating hypotheses
...
Selvin and Stuart (1966), therefore, designate three practices of survey analysts:
1
...
The process of searching through a body of data and looking at many relations
in order to find those worth testing (that is, there are no pre-designated hypotheses)

2
...
The process of using the data to choose which of a number of pre-designated
variables to include in an explanatory model

3
...
The process of testing from the data all of a pre-designated set of hypotheses

This investigative approach is reasonable for basic research but may not be practical for
decisional research
...
Rarely can the decision maker afford the luxury of dredging
through the data to find all of the relationships that must be present
...


MAKING INFERENCES
Testing hypotheses is the broad objective that underlies all decisional research
...
Often,
however, we cannot measure everyone in the population but instead must estimate the population
using a sample of respondents drawn from the population
...
Thus, in both estimation and hypothesis
testing, inferences are made about the population of interest on the basis of information from a
sample
...
Smith and Gerald S
...
Even in these cases, however, we would want to know about the
underlying associated variables that influence preference, purchase, or use (color, ease of
opening, accuracy in dispensing the desired quantity, comfort in handling, etc
...
In yet other case studies, it
might be necessary to analyze the relationships between the enabling or situational variables that
facilitate or cause behavior
...

The Relationship Between a Population, a Sampling Distribution, and a Sample
In order to simplify the example, suppose there is a population consisting of only five
persons
...
The frequency
distribution of the population is shown in the bar chart of Figure 11
...

Figure 11
...


Now that we know the “parameters” of the population, we will consider the sampling
distribution for our example data
...
In our example, the population is not large, and we will assume a sample of size n = 2
...
The sampling distribution of means and standard errors are shown in Table 11
...


269
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Table 11
...
Also
understand that the population mean
4 is always equal to the mean of the sampling
distribution of all possible sample means (
= 4 )
...
We must, therefore compute statistics from the sample ( x and s) and make
inferences about the population , and using the sample information
...
The
relevant statistics for the first of these samples having the values of (1,2) are:

270
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Given this x and
, we can now estimate with a given probability, the intervals that
give a range of possible values that could include , the population mean
...
5 ± 6
...
5) or -1
...
655
95% = 1
...
71 (
...
855 to 7
...
5 ± 31
...
5) or -14
...
41
Thus, we could state that we are 99% confident that the population mean would fall
within the interval –14
...
41
...
0
...
As
the sample size increases, the numbers become larger, and gradually approximate a standard
normal distribution
...
2
...
2
shows the relationship between the population, the sample, and the sampling distribution while
Part II illustrates the impact of sample size on the shape of the sampling distribution for
differently shaped population distributions
...
2 Population, Sample, and Sampling Distribution

Part I

271
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Part II

Acceptable Error in Hypothesis Testing

A question that continually plagues analysts is, what significance level should be used in
hypothesis testing? The significance level refers to the amount of error we are willing to accept
in our decisions that are based on the hypothesis test
...

In hypothesis testing the sample results sometimes lead us to reject H0 when it is true
...
On other occasions the sample findings may lead us to accept H0 when it
is false
...
The nature of these errors is shown in Exhibit 11
...

The amount of type I error, , we are willing to accept should be set after considering (a)
how much it costs to make such an error and (b) the decision rule used
...
There are substantial
differences in costs of errors in research conducted for exploratory purposes and research
conducted to make a decision where large financial investments are made
...

A tradition of conservatism exists in basic research and has resulted in the practice of
keeping the Type I error at a low level (
...
01)
...
The basic researcher typically assigns higher costs to a
Type I than to a Type II error
...
Smith and Gerald S
...
3

Types of Error in Making a Wrong Decision

There are two types of error that result from a mismatch between the conclusion of a research study
and reality
...
There are also two possibilities for the true situation: H0 is true or H0 is false
...


A Type I error occurs when we incorrectly conclude that a difference exists
...

2
...

3
...
This is equal to the area under the normal curve less the
4
...
Increasing sample size
preferred option for most market researchers
...
The cost of missing a market entry by not producing a product and foregoing gain (Type
II error) may be even greater than the loss from producing a product when we should not (Type I
error)
...
Of course not all
decision situations have errors leading to such consequences
...

In decisional research, the acceptable level of significance (i
...
, the specification of a
significance level or ) should be made by the client or the manager for whom the study has
been conducted
...
Smith and Gerald S
...
This means that the researcher should merely report the level
at which significance occurs, letting the manager decide what this means
...
05
...

Power of a Test

The power of a hypotheses test is defined as 1- , or 1 minus the probability of a Type II
error
...

The power of a statistical test is determined by several factors
...
This can
be controlled by increasing
...

In Figure 11
...
In testing the equality of the two distributions, we first identify
, which is here set to
...
(The one tail probably is
...
65 )
...
2, we observe
that for the second sampling distribution, power is the area to the right of µ0 + 1
...
e
...

Figure 11
...
65

x

x

If were decreased to, say, µ0 + 1
...

The second method of increasing power is to increase sample size
...

In decisions that imply a greater cost for a type II error (missing a difference that does in
fact exist), the analyst may want to consider expanding the probability of a type I error ( ) to
...
15, or even
...
Exhibit 11
...


SELECTING TESTS OF STATISTICAL SIGNIFICANCE
Up to this point, we have considered data analysis at a descriptive level
...
In many cases
this involves testing hypotheses concerning tests of group means
...
Smith and Gerald S
...
Often however, we have only nominal or loosely ordinal data
and we are not able to meet the rigid assumptions of a parametric test
...
The 2 statistic is from
the family of non-parametric methods
...
, 2)
...
These techniques are appropriate
for many Marketing applications where measurement is often at an ordinal or nominal level
...
The one that is appropriate for
analyzing a set of data depends on: (1) the level of measurement of the data, (2) the number of
variables that are involved, and for multiple variables, how they are assumed to be related
...
Corresponding to this spectrum of available information is an array of
non-parametric and parametric statistical techniques that focus on describing (i
...
, measures of
central tendency and dispersion) and making inferences about the variables contained in the
analysis--i
...
, tests of statistical significance (as shown in Table 11
...
A useful reference for
selecting an appropriate statistical technique is found in the Qualtrics Survey University
...
5

Selected Parametric and Nonparametric Univariate Analyses

What type of analysis do you want?
Measure of central tendency

Nominal
Mode

Level of Measurement
Ordinal
Interval (Parametric)
Median
Mean

Measure of Dispersion

None

Percentile

Standard Deviation

Binomial test

Kolmogorov-Smirnov
One-sample test

t-test

One-sample runs test

Z-test

One-Sample test of statistical
significance

EXHIBIT 11
...
But,
except in rare instances, very little has been said about research failure, or research that leads to
incorrect conclusions
...
The flaw may be inadequate
statistical power
...
By doing so, we automatically incur a high risk of Type II errormistaking a real difference for chance variation
...


275
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Power was hot stuff in the 1980s
...

2
...


The concept is more complicated than statistical significance or confidence limits
...

Numerical objectives must be specified before a research budget is fixed
...
Management, even if it has a target in mind, is often
unwilling to do so
...
Suppose also that (unknown to us) Option X if superior to Y by 10
percentage points, say 30% strong buying interest vs
...

If we use two samples of 200 each, how likely will our study identify X as being
significantly superior, using a 95% significance criterion? The likelihood of that correct result is only
68%, just a shade better than 2:1, in effect a 32% risk of research failure
...

Setting the desired levels of protection against the two types of error should not be a
rote process
...
By “testing” for statistical insignificance
one asks, “Is this result so likely to occur by chance that we should ignore it?” Is statistical
insignificance high enough to dismiss the finding?
But, as pointed out by one observer, there is no scientific, objective way to determine
what is “high enough” (Semon, 1999)
...
That’s the same
as saying the likelihood of a difference as large as observed occurring by chance is at odds of 1:3
...
But, the level that such
odds should have is a matter of personal preference and reflects one’s risk taking philosophy
...
Why do we need 95% protection against Type I
error? If we lower our significance criterion to 90%, the sample size requirement drops to 300
...
For example, for normally
distributed variables, we assume normality and homogeneity of variances
...
This process of making inferences from the
sample to the population's parameters is called parametric analysis
...
Parametric methods rely almost exclusively on either
276
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

interval or ratio scaled data
...
When data do not meet the rigorous assumptions of
parametric method, we must rely on non-parametric methods which free us of the assumptions
about the distribution
...
Other non-parametric methods that use an ordinal measurement scale test for the
ordering of observations in the data set
...
Often times, the researcher will find
that the same conclusion regarding significance is made when data are analyzed by a parametric
method and by its “corresponding” non-parametric method
...
In the next chapter bi-variate parametric and nonparametric analyses are presented
...

Univariate Analyses of Parametric Data
Marketing researchers are often concerned with estimating parameters of a population
...
Very often, the means, proportions and variances
are the summary measures of concern
...
These comparisons
involve a single variable
...
These issues are
discussed in more depth by Mohr (1990)
...
The
confidence interval is a range of values with a given probability (
...
99, etc
...

For example, assume we have a normally distributed population with population mean µ
and a known population variance 2
...
This
single item is an estimate of µ, the population mean
...
This normal distribution permits us to estimate the probability
associated with various intervals of values of X
...
96
about 95% of the area under the normal probability curve is within this range
...
We
compute this range for the sample problem that follows
...
00 per month, and the known
population standard deviation is =10
...
The 95% confidence interval about the population
mean is computed as:

277
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Lower limit population mean
µ - 1
...
96)(10)
32 - 19
...
4

upper limit

Thus, we expect that 95% of all household expenditures will fall within this range, as shown in
Figure 11
...

Figure 11
...
025


...
40

µ x = $32
...
96

Reject H o

$51
...
96

Accept H o

x
z
Reject

Ho

If we expand this analysis to construct a confidence interval around the mean of a sample
rather than the mean of a population we must rely on a sampling distribution to define our
normal distribution
...
Recall that the population mean µ is also the mean of the normally distributed sampling
distribution,
s/ )) describes the confidence interval for the sampling distribution
...
It is interesting to note
that because n, the sample size, is included in the computation of the "standard error," we may
estimate the population mean with any desired degree of precision, simply by having a large
enough sample size
...
The z statistic describes probabilities of the normal distribution and is the
appropriate tool to test the difference between µ, the mean of the sampling distribution, and, the
sample mean when the population variance is known
...

(2) The population must be normally distributed
...

(3) The data must be at least interval scaled
...

278
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

1
...
Any
observed difference is due solely to sample variation
...
05)
...

(5) If the probability of the observed differences having occurred by chance is greater than
the alpha used, then H0 cannot be rejected and it is concluded that the sample mean is
drawn from a sampling distribution of the population having mean µ
...
The sample has a mean of =$30
...
00
...
00 equals the
sample mean =30
...

Steps:

The results are shown in Figure 11
...

Figure 11
...
Smith and Gerald S
...
With this variance estimate, we compute the t
statistic
...
00, when the sample of 225 shoppers shows a mean, , of $30
...
00
...
00, df = n-1=224) =
...
00) =
...
96 s/ )= 1
...
176 = 30
...
Figure 11
...

Figure 11
...
05


...
80
x

32
33
...
824
-2
...
96

0

+1
...
The answer is relatively simple
...
It is the same regardless of the sample that is drawn, as long as the sample is of size n
...
Because the variance of the
sample was less than the variance of the population (9 vs
...


280
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Unlike the z-distribution, the t is a family of distributions, each having a different shape,
depending on the degrees of freedom (see Figure 11
...
The probability for a given value of t
varies across distributions
...
" Unfortunately, the concept of degrees of freedom is inadequately defined
...
A second
definition is that “degrees of freedom” is a measure of how much precision an estimate of
variation has
...
This concept seems to something that “we all know what it is, but cannot
precisely define it
...

In univariate analyses, the available degrees of freedom are n-1
...
To explain further, suppose we
have a sample with size n = 5 in which the mean value of some measure is calculated to be zero
...
Each time we estimate a parameter we lose
one degree of freedom, while n - 1 are free to estimate that parameter
...
8

The t-distributions for df = 1, df = 4 and df = 10

As n becomes large (
reached at the limit (n = )
...

The population must be normally distributed
...

The data must be at least interval scaled
...

281

Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Univariate Analysis of Categorical Data:
The Chi-Square Goodness of Fit Test
2
Chi-square analysis ( ) can be used when the data identifies the number of times or
frequency that each category of a tabulation or cross tabulation appears
...


Determining the significance of sample deviations from an assumed
theoretical distribution; that is, does a certain model fit the data
...


2
...
This is typically called a test of
independence
...


The procedure involved in chi-square analysis is quite simple
...
These
theoretical frequencies may result from application of some specific model of the phenomenon
being investigated--relationship 1 above
...

In either case we compute a measure (chi-square) of the variation between actual and
theoretical frequencies, under the null hypothesis that there is no difference between the model
and the observed frequencies
...
If the measure of
variation is "high," we reject the null hypothesis at some specified alpha risk
...

Single Classification

Suppose that we are interested in how frequently a sample of respondents selects each of
two test packages
...
Test package B is the same package but,
rather than an attached coupon, it contains an attached premium (a ballpoint pen) that the
respondent may keep
...

The frequency of choice is presented in the first column (labeled "Observed") of Table
11
...
We see that 63 respondents out of 100 select package A
...
Given that the researcher's model would predict an estimated
frequency of 50-50, are the observed frequencies compatible with this theoretical prediction? In
chi-square analysis we set up the null hypothesis that the observed (sample) frequencies are
consistent with those expected under application of the model
...
Assume that there are k categories and a random sample
of n observations; each observation must fall into one and only one category
...
Smith and Gerald S
...
6
Observed vs
...
38
B
37
50
-13
169/50 = 3
...
67

In the above one-way classification problem, the statistic is approximately distributed as
chi square with k-1 degrees of freedom
...
Table A
...

In Table A
...
05 and k-1=1 is 3
...
If the null
hypothesis is true, the probability of getting a chi-square value greater then 3
...
05
...
76 (see Table 11
...
9)
...
Sample size is important to the quality of the approximation and the power of
the test
...
3 will
not be a good approximation
...

Figure 11
...
Smith and Gerald S
...
The univariate test of proportions, like the univariate test of
means, compares the population proportion to the proportion observed in the sample
...
225 respondents are surveyed at a local shopping mall
...
The manager needs a 90% favorability rate
...
05 with z = 1
...

CI =
...
96(
...
96(
...
856 <
...
944

In this example the manager could (in a statistical sense) claim that 87% approval is no
different that 90% approval with sampling variation
...


SUMMARY
Chapter 11 has introduced the basic concepts of formulating hypothesis testing and
making statistical inference in the context of univariate analysis
...

A hypothesis is a statement that variables (measured constructs) are related in a specific
way
...

Statistics are based on making inferences from the sample of respondents to the
population of all respondents by means of a sampling distribution
...

284
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

When testing hypotheses, the analyst may correctly identify a relationship as present or
absent, or may commit one of two types of errors
...
A Type II Error occurs when we accept a
false H0 (there is a difference, but we find that none exists)
...

Selecting the appropriate statistical technique for investigating a given relationship
depends of the level of measurement (nominal, ordinal, interval) and the number of variables to
be analyzed
...
non-parametric analyses depends on the analyst's
willingness to accept the distributional assumptions of normality and homogeneity of variances
...
The t-test
was demonstrated as a parametric test for populations of unknown variance and samples of small
size
...
The observed frequencies were compared to an
expected distribution
...
Chapter 12 expands this discussion to
include non-parametric and parametric analyses involving two variables
...
Finally, in Chapters 13 and 14,
we focus on multivariate analyses, involving dependence, and interdependence of associative
data
...

However, without higher levels of measurement and their associated characteristics of central
tendency dispersion and rates of change, the ability to investigate these more complex
relationships is not possible
...
Smith and Gerald S
...
M
...
, Davidson, T
...
, O’Malley, P
...
, and Rodgers, W
...
(1981), A Guide
for Selecting Statistical Techniques for Analyzing Social Science Data
...
(2003)
...
Thousand Oaks, CA: Sage
Publications, Second Edition
Gibbons, J
...
(1993)
...
Newbury Park, CA: Sage
Publications
Labovitz, S
...
“Criteria for Selecting A Significance Level: A Note on the Sacredness of

...

Lewis-Beck, M
...
(1995)
...
Thousand Oaks, CA: Sage
...
B
...
Understanding Significance Testing
...
(1995)
...
Los Angeles: Pyrczak
Publishing
Selvin, H
...
(1966)
...
T
...
“When Determining Significance, Let Clients Decide the Risks,” Marketing
News October 9, 15
Semon, T
...
(1999)
...
T
...
:Keep Ignoring Statistical Power at Your Own Risk,” Marketing news, 24,
181 (September 30), 21

286
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 12
BIVARIATE DATA ANALYSIS
Analysis can be viewed as the categorization, aggregation, and the manipulation of data
to obtain answers to research question underlying the research project
...
The process of interpretation involves taking the results of
analysis, making inferences relevant to the research relationships studied, and drawing
managerially useful conclusions about these relationships
...
A large part
of the rest of the chapter will focus on methods for analyzing how the variation of one
variable is associated with variation in other variables
...
We then show how various insights can be obtained as one
goes beyond two variables in a cross-tabulation to three (or more) variables
...

Bivariate Cross-Tabulation
Bivariate cross-tabulation represents the simplest form of associative data
analysis
...

Data of this type is called categorical, since each variable is assumed to be nominalscaled
...
In fact, it is the single most widely used bivariate
technique in applied settings
...
376):
1
...

2
...

3
...

4
...


287
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The entities being cross-classified are usually people, objects, or events
...
Excellent discussions
of ways to analyze cross-tabulations can be found in Hellevik (1984) and Zeisel (1957)
...
At the
very least, row or column percentages (or both) are usually computed
...
In crosstabulation, percentages serve as a relative measure indicating the relative size of two or
more categories
...
Yet its simplicity
of calculation is sometimes deceptive, and the understanding of its purpose is frequently
insufficient to ensure sound application and interpretation
...

The sources of problems in using percentages are largely in identifying the
direction in which percentages should be computed, and in knowing how to interpret
percentage of change
...
Let us assume that KEN’s Original, a
small regional manufacturer of salad dressings, is interested in testing the effectiveness of
spot TV ads in increasing consumer awareness of a new brand—called Life
...
The test
area receives five 15-second television spots per week over an eight-week period,
whereas the control area receives no spot TV ads at all
...
)
Assume that four independent samples (of telephone interviews) were conducted
as before and after tests in each of the areas
...
If Life was mentioned, it
was assumed that this constituted consumer awareness of the brand
...
This common
occurrence in surveys (i
...
, variation in sample sizes) increases the value of computing
percentages
...
1 shows a set of frequency tables that were compiled before and after a
TV ad for Life Salad Dressing was aired
...
1 would be hampered
if the data were expressed as raw frequencies and different percentage bases were
reported
...
1 shows the data, with percentages based on column and
row totals
...
In cases where this distinction is clear, the rule is to
compare percentages within levels of the dependent variable
...
Smith and Gerald S
...
1, the control variable is the experimental area (test versus control)
and the dependent variable is awareness
...
We note that before the spot TV campaign
the percentage of respondents who are aware of Life is almost the same between test and
control areas: 42 percent and 40 percent, respectively
...
The
small increase of 2 percentage points reflects either sampling variability or the effect of
other factors that might be serving to increase awareness of Life in the control area
...
We note that 61 percent of the aware group (before
the spot TV campaign) originates from the test area; however, this is mainly a reflection
of the differences in total sample sizes between test and control areas
...
This may be
erroneously interpreted as indicating that spot TV in the test area depressed awareness in
the control area
...

It is not always the case that one variable is clearly the independent or control
variable and the other is the dependent or criterion variable
...
Indeed, cases often arise in which each of the variables in
turn serves as the independent and dependent variable
...
1

Before Spot TV

After Spot TV

Aware
250
61%
42%

Total
Before TV Spot

350
59%
58%

600
60%

160
39%
40%

Test Area
Freq
...

Col %
Row %

Not
Aware

240
41%
60%

400
40%

410
41%

590
59%

1000
100%

Total

Area
Test Area
Freq
...

Col %
Row %
Total
After TV Spot

Aware

Not
Aware

Total
Area

330
67%
66%

170
44%
34%

550
57%

160
33%
42%

220
56%
58%

380
43%

490
56%

390
44%

880
100%

Interpretation of the Percentage Change

A second problem that arises in the use of percentages in cross-tabulations is the choice
of which method to use in measuring differences in percentages
...
The absolute difference in percentages
2
...
The percentage of possible change in percentages
The same example can be used to illustrate the three methods
...
Smith and Gerald S
...
2 shows the percentage of respondents who are aware of Life before and after
the spot TV campaign in the test and control areas
...
The increase for the testarea respondents was 24 percentage points, whereas the control-area awareness increased
by only 2 percentage points
...
2
Test area
Control area

Before the
Campaign
42%
40%

After the
Campaign
66%
42%

Relative Percentage Increase

The relative increase in percentage is [(66 – 42)]/42] × 100 57 percent and [(42
– 40)/40] × 100 5 percent, respectively, for test- and control-area respondents
...
The increase actually registered is 24 percentage points, or 100(24/58) 41
percent of the maximum possible
...

In terms of the illustrative problem, all three methods give consistent results in the
sense that the awareness level in the test area undergoes greater change than that in the
control area
...

The absolute-difference method is simple to use and requires only that the
distinction between percentage and percentage points be understood
...
The percentage-of-possible-difference method takes
cognizance of the greater difficulty associated with obtaining increases in awareness as
the difference between potential-level and realized-level decreases
...

Introducing a Third Variable into the Analysis
Cross-tabulation analysis to investigate relationships need not stop with two
variables
...
In
some cases, it may show that the two variables are related even though no apparent
relationship exists before the third variable is introduced
...

Consider the hypothetical situation facing MCX Company, a specialist in
telecommunications equipment for the residential market
...
Several months after the introduction, a telephone survey was taken in which
respondents in the test area were asked whether they had adopted the innovation
...
3)
...
Smith and Gerald S
...
3
Frequency
Adopters
Number of Cases
Column %
Row %
Non Adopters
Number of Cases
Column %
Row %
Total
Percentage

Adoption—Percentage by Gender and Age
<35 Yrs

Men
35 Yrs

Total %

<35 Yrs

Women
35 Yrs

Total %

100
50%
62
...
5%

160
40%

11
11%
55%

9
9%
45%

20
10%

100
50%
41
...
3%

240
60%

89
89%
49
...
6%

180
90%

200
50%

200
50%

400
100%

100
50%

100
50%

200
100%

One of the variables of major interest in this study was the age of the respondent
...
Accordingly, the market analyst decides
to cross-tabulate adoption and respondent age
...
Table 12
...
It seems that the total sample of 600 is split evenly between those who are
under 35 years of age and those who are 35 years of age or older
...

Analysis and Interpretation
The researcher is primarily interested in whether this finding differs when gender
of the respondent is introduced into the analysis
...

Table 12
...
In the case of men, 50 percent of the younger men adopt compared with only 30
percent of the older men
...
Even here, however, younger women show a slightly higher percentage of
adoption (11 percent) than older women (9 percent)
...
However, what can now be said is
the following: If the respondent is a man, the differential effect of age on adoption is
much more pronounced than if the respondent is a woman
...
1)
...
The relative width of the bars denotes the relative size of
the categories—men versus women—representing the third variable, gender
...


291
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 12
...
Furthermore, the size of the difference depends on the gender of the
respondent: Men display a relatively higher rate of adoption, compared with women, in
the younger age category
...
In the example presented, adoption and age exhibit initial association; this association
is still maintained in the aggregate but is refined by the introduction of the third variable,
gender
...
Adoption and age do not appear to be associated
...
In the two-variable cases, men and
women exhibit opposite patterns, canceling each other out
...
It goes almost without saying that the introduction of a
third variable can often be useful in the interpretation of two-variable cross-tabulations
...
Association of two or more variables
does not imply causation, and this statement is true regardless of our preceding efforts to
refine some observed two-variable association through the introduction of a third
variable
...
g
...
However, somewhere along the line, a problem arises in maintaining an
adequate cell size for all categories
...
A further problem, independent
of sample size, concerns the high degree of complexity of interpretation that is introduced
by cross-tabulations involving four or more variables
...

292
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

As noted in Table 12
...
In many applications,
however, the criterion variable will have more than two categories
...


BIVARIATE ANALYSIS: DIFFERENCE BETWEEN SAMPLE GROUPS
Marketing activities largely focus on the identification and description of market
segments
...
The key component of each of these variables is the ability
to group respondents into market segments
...
These analyses are often differences between respondent groups
...

We begin with what is perhaps the most used test of market researchers: cross-tabulation
...
First, we discuss the t-test of
differences in means of two independent samples, and then we look at one-way analysis
of variance (ANOVA) that tests for differences in means for k groups
...
These
are but a few of the possible parametric and nonparametric analyses that could be
discussed (see Table 12
...

Bivariate Cross-Tabulation
The chi-square statistic in a contingency table analysis is used to determine if the
observed associations between the variables in the cross-tabulation are statistically
significant
...
In
2
conjunction with the cross-tabulation we will introduce the chi-square statistic, , to
determine the significance of observed association in cross-tabulations involving two or
more variables
...

Cross-tabulation represents the simplest form of associative data analysis
...
We know that such data are called categorical because each variable is
assumed to be only nominal-scaled
...
In fact, it is the
single most widely used bivariate technique in applied settings
...
Smith and Gerald S
...
4

Selected Nonparametric Statistical Tests for Two-Sample Cases
Level of Measurement

What type of analysis
do you want?
Two-sample related
samples
Independent samples

Nominal

Ordinal

McNemar test for the
significance of changes

Sign test
Wilcoxon matched-pairs
signed-ranks test

Fischer exact probability
test

Median test
Mann-Whitney U test
Kolmogorov-Smirnov
two-sample test
Wald-Wolfowitz runs test

2

X test for two
independent samples

Interval / Ratio

t-test
One-way ANOVA

In marketing research, observations may be cross-classified, as when we are
interested in testing whether occupational status is associated with brand loyalty
...
The data for our
hypothetical problem appear in Table 12
...

A total of four columns, known as banner points, are shown
...
Professional cross-tabulation software will output multiple side by side
tables that join multiple variables on the column banner points such that loyalty and
another variable such as usage occasion could be analyzed simultaneously
...
The data suggests that a
relationship exists, but is the observed association a reflection of sampling variation, or
can we conclude that a true relationship exists?
Table 12
...
1%)

Highly Loyal
30 (30
...
01

Moderately Loyal
42 (34
...
86

Brand Switchers
18 (25
...
17

Blue Collar

14 (22
...
93

20(24
...
86

31 (18
...
68

65 (28
...
4)
2
...
4)

...
2)
1
...
6%)

78 (33
...
8%)

65 (28
...
08

White-collar

Total

In analyzing the problem by means of chi-square, we make use of the marginal
totals (column and row totals) in computing theoretical frequencies given that we
hypothesize independence (no relationship) between the attributes loyalty status and
occupational status
...
5 that 33
...
If possession of this characteristic is independent of
occupational status, we would expect that 33
...
e
...
5) would be highly loyal
...
8
294
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

percent (87/320) of the 90 = 34
...
3 percent (65/320)
of the 90 = 25
...
In a similar fashion we can compute
theoretical frequencies for each cell on the null hypothesis that loyalty status is
statistically independent of occupational status
...
)
The theoretical frequencies (under the null hypothesis) are computed and appear
in parentheses in Table 12
...
The chi-square statistic is then calculated (and shown in the
table) for each of the data cells in the table using the observed and theoretical
frequencies:

where fi actual observed frequency, Fi theoretical expected frequency, and k number
of cells (r x c)
...
In
general, if we have R rows and C columns, the degrees of freedom associated with the
chi-square statistic are equal to the product
If we use a significance level of 0
...
488
...
08 is greater than the table value of 11
...

A correction factor must be applied to the formula for chi-square when the
number of observations in a cell is less than 10 (or where a 2 × 2 contingency table is
involved),
...
This correction factor adjusts for the use
of a continuous distribution to estimate probability in a discrete distribution
...
No new principles are involved
...
First, chi-square analysis deals with counts (frequencies) of
data
...
Second, the technique assumes that the observations are drawn
independently
...

To assess the nature of the relationship, the researcher must look at the table and
indicate how the variables appear to be related—a type of eyeball approach
...
When variables are ordered or loosely ordered, a
pattern can sometimes be observed by marking cells with higher than expected observed
frequencies with a (+) and those with lower than expected observed frequencies with a (–
), or even graphing the deviations from expected values, or cell X2 values using a 3dimensional contour map
...
Smith and Gerald S
...
In addition, many studies go beyond estimation and compare
such population parameters by testing hypotheses about differences between them
...
Our
concern at this point is with differences in means and proportions
...

Standard Error of Differences
Here we extend the topic of sampling distributions and standard errors as they
apply to a single statistic - to cover differences in statistics and show a traditional
hypothesis for differences
...
If the population standard deviations, i, are not known, then the estimated
standard error becomes

For relatively small samples the correction factor Ni/ni–1 is used and the resulting
formula for the estimated standard error is

Of course, these formulas would be appropriate for use in the denominator of the
t-test
...
Specifically, for large samples,

For small samples, the correction factor is applied, resulting in

296
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

This estimate would again be appropriate for use in the denominator of the Z-test
of proportions
...
Samples must be independent
...
Individual items in samples must be drawn in a random manner
...
The population being sampled must be normally distributed (or the sample
must be of sufficiently large size)
...
For small samples, the population variances must be equal
...
The data must be at least interval scaled
...

1
...
g
...

2
...
05 or other value)
...
A Z value is calculated by the appropriate adaptation of the Z formula
...
The probability of the observed difference of the two sample statistics having
occurred by chance is determined from a table of the normal distribution (or
the t distribution, interpreted with [nA + nB – 2] degrees of freedom)
...
If the probability of the observed difference having occurred by chance is
greater than the alpha risk, the null hypothesis of no difference is not rejected;
it is concluded that the parameters of the two universes are not significantly
different
...
In an
applied setting, there are times when the level at which significance occurs
(the alpha level) is reported and management decides whether to accept or
reject
...
Smith and Gerald S
...
Let us assume we
have conducted a survey of detergent and paper goods purchases from supermarkets
among urban (population A) and rural (population B) families (see Table 12
...

Table 12
...
00

SA = $10
...
80

SB = $9
...
20 difference in means caused by sampling variations?”
We proceed as follows
...
We
assume the alpha risk is set at
...
Since a large sample test is called for, the Z value is
calculated using the separate variances estimate of the standard error of differences in
means:

The Z value is then determined to be

The probability of the observed difference in the sample means having been due
to sampling is specified by finding the area under the normal curve that falls to the right
of the point Z
1
...
Consulting a table of the Cumulative Normal Distribution, we
find this area to be 1
...
9382 0
...
Since this probability associated with the
observed difference ( p 0
...
In a decision setting, however, the manager would
have to determine whether this probability (0
...

As stated previously, and discussed in the previous chapter, often there is no
preset alpha and decisional considerations require that the manager determines the
meaning of the reported alpha
...
With these data we calculate t as follows:

298
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The critical value of t is obtained from a table of percentiles for the t distribution
...
05, we determine the critical value of t for (nA nB 2) 25 degrees of
freedom to be 1
...

Since the calculated t of 1
...
708, we cannot reject the hypothesis of no difference in
average amount spent by the two types of families
...
4
...
The
formulas for calculating the test statistics differ, however
...
2

t-Distribution

Testing the Means of Two Groups: The Independent Samples t-Test
The t-distribution revolutionized statistics and the ability to work with small
samples
...

For most market research applications it is difficult to justify the Z-test’s assumed
knowledge of µ and
...
The t-test is a widely used market research
statistic to test for differences between two groups of respondents
...
Most computer programs
recognize two versions of this statistic
...

The second method, called the pooled variance estimate, computes an average for
the samples that is used in the denominator of the statistic:

The pooling of variances is a simple averaging of variances that is required when
(1) testing for the same population proportion in two populations or (2) testing the
difference in means between two small samples
...
7 shows a sample output of t-tests from SPSS
...
Smith and Gerald S
...
Respondents who were males
2
...
It will be noted that the
sample sizes are approximately equal
...
In all three examples, the differences between separate-variance and pooledvariance analysis are very small
...
9 also shows an F statistic
...
Using this test the researcher knows which t-test output to
use—equal variances assumed or equal variances not assumed
...
7

Selected Output From PASW (SPSS) t-Test

Testing of Multiple Group Means: Analysis of Variance
Analysis of variance (ANOVA) is a logical extension of the independent groups ttest methodology
...
ANOVA is a general set of methodologies
that handle many different types of research and experimental designs
...
The respondent receives a combination of the
treatment levels from the different factors and the response is measured
...

Much experimental work has been conducted in medicine and agriculture
...
The
variables influencing the results are called experimental factors
...
Smith and Gerald S
...
Control variables in this application might
include seed type, fertilizer type, fertilizer dosage, temperature, moisture, and many other
variables thought to influence production
...
Marketing research experiments have control variables that are
certainly different from agricultural experiments, but the principles are the same
...
Here we discuss the
two most basic forms of the ANOVA methodology: the one-way ANOVA and the twofactor ANOVA designs (see Exhibit 12
...

EXHIBIT 12
...

A simple one-way ANOVA to investigate this relationship might compare three messages:
Advertising message A
A

Advertising message B
B

Advertising message
C
C

Two-factor ANOVA includes a second factor, possibly the type of advertisement (magazine or
TV):
Message (A)
Message (B)
Message (C)
Magazine Ad (M)
MA
MB
MC
TV Ad (T)
TA
TB
TC
Each of the cells in this matrix would contain an average interest rating for the measures taken for
the particular combination of message and media
...
This is not strictly the case
...

The ANOVA Methodology
The appropriateness of the title analysis of variance comes from the method’s
ability to explain the variation in responses to the various treatment combinations
...
2, which presents an
example regarding responses to messages
...
Smith and Gerald S
...
2 Example of ANOVA Methodology
Using the single-factor ANOVA design for the problem described in Exhibit 12
...

1
...


Total Sum of Squares:
Grand mean = 4
...
Between-treatment sum of squares: The means of the factor levels (messages A, B, and C) are
computed, followed by the deviation of the factor level means from the overall mean, weighted
2
by the number of observations
j- )
3
...


302
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Thus, an observation may be decomposed into three terms that are additive, and each
explains a particular type of variance:

The overall mean is constant and common to all observations: The deviation of a group
mean from the overall mean represents the effect on each observation of belonging to that
particular group; the deviation of an observation from its group mean represents the effect on that
observation of all variables other than the group variable
...
This is the F statistic that indicates the strength of the grouping factor
...
Conversely, if the three sample means
were very close to each other, the between-samples sum of squares would be close to
zero and we would conclude that the population means are the same, once we consider
the variability of individual cases within each sample group
...
Smith and Gerald S
...
This is exactly the same
assumption as was made for the t-test
...
Still, the foregoing remarks represent the basic
ANOVA idea for comparing between- with within-sample variability
...

Suppose a new product manager for the hypothetical Friskie Corp
...
The product manager has
been able to secure the cooperation of a store manager to run an experiment involving
three levels of shelf height (knee level, waist level, and eye level) on sales of a single
brand of dog food, which we shall call Snoopy
...
But what shall we use for our unit
of time? Sales of dog food in a single store may exhibit week-to-week variation, day-today variation, and even hour-to-hour variation
...

Assume that we have agreed to change the shelf-height position of Snoopy three
times per day and run the experiment over eight days
...
Furthermore, since our primary emphasis is on
explaining the technique of analysis of variance in its simplest form (analysis of one
factor: shelf-height), we shall assign the shelf heights at random over the three time
periods per day and not design an experiment to explicitly control and test for within-day
and between-day differences
...
8
...
If we look at mean sales by each level of shelf height, it appears as
though the waist-level treatment, the average response to which is X2 90
...
However, we note that the last
observation (93) under the eye-level treatment exceeds the waist-level treatment mean
...

Now we shall show what happens when one goes through a typical one-way
analysis-of variance computation for this problem
...
Smith and Gerald S
...
8

Sales of Snoopy Dog Food (in Units) by Level of Shelf Height

Table 12
...
Had the experimenter used
an alpha risk of 0
...
A table of F ratios is found in most statistics and research texts
...
9

Analysis of Variance – Snoopy Dog Food Experiment

Note that Table 12
...

For example, the total sum of squares is given by

305
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

This is the same quantity that would be obtained by subtracting the grand mean of
85
...
This mean-corrected sum of squares is equivalent to the type of formula used
earlier in this chapter
...
6 (2, 21 df) with the table value of 4
...
05)
...
6 4
...

The important consideration to remember is that, aside from the statistical
assumptions underlying the analysis of variance, the variance of the error distribution
will markedly influence the significance of the results
...
As we know, an increase in
sample size can reduce this experimental error
...

Follow-up Tests of Treatment Differences
The question that now must be answered is the following: Which treatments
differ? The F-ratio only provides information that differences exist
...
Because of our previous discussion of the t-test, we will not discuss these tests
in detail
...
These test statistics (which
include techniques known as the LSD (Least Significant Difference), Bonnferoni’s test,
Duncan’s multiple range tests, Scheffe’s test, and others) control the cumulative
probability that a Type I error will occur when a series of statistical tests are conducted
...
05 probability of a Type I
error, then in a series of 20 such tests we would expect one (20 *
...
These tests typically
are options provided by the standard statistical packages, such as the PASW (SPSS)
program Oneway
...
Our concern is with the nature of the associations
between the two variables and the use of these associations in making predictions
...
As an initial step in studying the
relationship between the X and Y variables, it is often helpful to graph this relationship in
a scatter diagram (also known as an X-Y plot)
...
3
...
Smith and Gerald S
...
The Pearson
correlation coefficient is commonly used this purpose and is defined by the formula

where n pairs of (Xi, Yi) values provide a sample size n, and X, Y, SX, and sY represent the
sample means and sample standard deviations of the X and Y variables
...

These extreme values indicate perfect positive and negative linear correlations
...

The alternate formulation shows the correlation coefficient to be the product of
the Z scores for the X and Y variables
...
The Z scores will be centered as a normally distributed
variable (mean of zero and standard deviation of one)
...
The original metric scales are replaced by a new abstract scale (called
correlation) that is the product of the two Z distributions
...
3

Scatter Diagrams

307
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Alternative formula:
Where

By continuing our digression one step further, we can show how the correlation
coefficient becomes positive or negative
...
When
both ZX and ZY are positive or both are negative, rXY is positive, as shown in Figure 12
...

When ZX is positive and ZY negative (or the opposite), a negative correlation will exist
...

Figure 12
...
Thus,
correlations of
...
7 are of exactly the same strength, but the relationships are in
opposite directions
...
Often, other factors underlie and are even responsible for the relationship
between the variables
...
We may not imply that Kraft sales directly caused fluctuations
in the national economy, or vice versa
...
In times of reduced family income,
macaroni and cheese is a low-price dietary substitute for more expensive meals
...
The data and plot appear in Figure 12
...

308
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

In order to calculate the correlation coefficient, we will reduce the previously
identified alternative formula to the basic computational formula consisting of sums for
the X and Y variables
...
10
...
5

Consumption Expenditure and Income Data

Table 12
...

Researchers routinely perform their analyses using Excel spreadsheets or computer
packages
...
Smith and Gerald S
...
Not surprisingly, the linear regression
model—as applied in either the bivariate (single predictor) or multivariate form (multiple
predictors)—is one of the most popular methods in the marketing researcher’s tool kit
...

The regression model has been applied to problems ranging from estimating sales
quotas to predicting demand for new shopping centers
...


Regression analysis in its simplest bivariate form involves a single dependent
(criterion) and a single independent (predictor) variable
...
In considering the use of either simple or multiple
regressions, the researcher is interested in four main questions:
1
...
If we can, how strong is the relationship; that is, how accurately can we predict
values of the criterion variable from values of the predictor (linear composite)?
3
...
Which predictor is most important in accounting for variation in the criterion
variable? (Can the original model be reduced to fewer variables, but still provide
adequate prediction of the criterion?)
The basic ideas of bivariate regression are most easily explained by a numerical
example
...

310
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Suppose that a marketing researcher is interested in consumers' attitudes toward
nutritional additives in ready-to-eat cereals
...

The researcher obtains consumers' interval-scaled evaluations of ten concept
descriptions using a preference rating scale that ranges from 1, dislike extremely, up to 9,
like extremely well
...
11 Preference Ratings of Ten Cereal Concepts Varying in Protein
(1)
Consumer
1
2
3
4
5
6
7
8
9
10
Total
Mean
Std
...


Figure 12
...
3
2
...
3
2
...
Figure 12
...
It appears that there is a direct relationship
between Y and X1
...
A scatter diagram is a
useful tool for aiding in model specification
...
3
...
Smith and Gerald S
...
3

Look at Your Data Before You Analyze

In deciding which type of regression approach to use, it is important that the
researcher know the shape of the interrelationship
...
Looking at this visually helps decide whether the
relationship is, or approximates being, linear or whether it has some other shape which
would require a transformation of the data—by converting to square roots or logarithms-or treatment as nonlinear regression (Semon, 1993)
...
e
...
A decision has to be made whether to retain
such outliers in the data set for analysis
...
Of course, what the analyst is seeking
is a regression function that has the best fit for the data, and this is typically based on
minimizing the squares of the distances between the actual and estimated values—the socalled least-squares criterion
...

But how do we find the numerical values of a and b? The method used in this
chapter is known as least squares as discussed in Exhibit 12
...
As the reader will recall
from introductory statistics, the method of least squares finds the line whose sum of
squared differences between the observed values Yi and their estimated counterparts i
(on the regression line) is a minimum
...
11
...
3)(4
...
3 )2
= 0
...

Having found the slope b, the intercept a is found from

312
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

a=Y - bX
= 4
...
886(4
...
491
leading to the linear function

ˆ
Y = 0
...
886X

This function is drawn with the scatter plot of points in Figure 12
...
It appears to
fit the plotted points rather well, and the model seems to be well specified (a linear rather
than arc linear or other form seems to fit)
...
Although leastsquares regression models do not need to assume normality in the (conditional)
distributions of the criterion variable, this assumption is made when we test the statistical
significance of the contribution of the predictor variable in explaining the variance in the
criterion (does it differ from zero?) With this in mind the assumptions of the regression
model are as follows (the symbols
a
and b:
1
...

Our particular sample we assume that each y value is drawn independently of all
others
...
Modern
versions of the model permit the predictors to be random variables, but their
distribution is not allowed to depend on the parameters of the regression
equation
...


The means of all of these normal distributions of Y lie on a straight line with

3
...
This (common) variance
does not depend on values assumed by the variable X
...
0), there is unexplained error in predicting Y
...
3 as the
difference between the values Yi and the regression line
model is expressed algebraically as:
Y = + X1 +
where
1=0
1

error term drawn independently from a normally distributed universe with
1

The nature of these assumptions is apparent in Figure 12
...
The reader should
note that each value of X has associated with it a normal curve for Y (assumption 1)
...

313
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

What if the dependent variable is not continuous? Exhibit 12
...
The analysis proceeds generally
as we are discussing it—the major change is that a transformation has been applied to the
dependent variable values
...
4

When to Use Logistic Regression

Data collected for customer satisfaction research provides a good illustration of when the
researcher should consider transformation of data
...
Many believe that customer satisfaction ratings
obtained on rating scales are not normally distributed, but are skewed toward higher scale
values (Dispensa, 1997)
...

Ultimately, overall a customer is either satisfied or not satisfied
...
Typically those customers who rate at the upper end of the scale, say 9
or 10 on a 10-point scale, are considered satisfied while all others are considered to be not
satisfied
...

A binary overall customer satisfaction variable follows the logistic distribution, thus allowing
for the use of logistic regression
...
Log-odds specify the direct association between the
independent variable and the dependent variable
...

Typically, logistic regression is used for multiple regression situations where there are two or
more independent variables
...
The key to its
being of value is the nature of the dependent variable, not the independent variable(s)
...
7 Two-Variable Regression Model--Theoretical

314
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

In constructing the estimating equation by least squares we have computed a
ˆ
regression line for a sample, and not for the population: Y = a + bX 1 where
theoretical model
...
6 for the specific
bivariate problem of Table 12
...

However, functional forms other than linear may be suggested by the preliminary
scatter plot
...
8 shows various types of scatter diagrams and regression lines for
the two-variable case
...
We note that the regression line passes through the
mean of each variable and that the slope b happens to be positive
...
In Panel II we note that there is residual
variation in Y, and, furthermore, that the slope b is negative
...
In this case the mean of Y is as
good a predictor as the variable X1 (the slope b is zero)
...
That is, no linear association is found (b = 0), even though a
curvilinear relationship is apparent from the scatter diagram
...
8 illustrates the
desirability of plotting one's data before proceeding to formulate a specific regression
model
...
8

Illustrative Scatter Diagrams and Regression Lines

Strength of Association

It is one thing to find the regression equation (as shown in Figure 12
...
Does the regression line
= a + bx (which uses X as a predictor) explain the variation in Y (predict Y)? To answer
this question, consider that the total variation in the Y variable may be divided into two
component parts: (1) variance that is explained by the regression line, and (2)
unexplained variation (residual)
...
9
...
Smith and Gerald S
...
9

(Yi

Scatter Diagram and Regression Line for Cereal Problem

(Y i Y ) 2 +

Y )2

Total variation or
Total Sum of
Squares Deviation
from the Mean

(Y i Y ) 2

Variation Explained
by the Regression or
Sum of Squares Due to
Regression

Variation
Unexplained by the Regression
...
This coefficient varies between 0 and 1 and
represents the proportion of total variation in Y (as measured about its own mean Y) that
is accounted for by variation in X1
...

If we were to use the average of the Y values (Y) to estimate each separate value
of Y, then a measure of our inability to predict Y would be given by the sum of the
n

squared deviations

( Y i - Y )2
...
In this case a measure of our inability to predict Yi is given by
i=1

can define r2 as a function of these two qualities:
n
2
r =1 -

i=1
n

ˆ
( Y i - Y I )2
...
However, if using the regression equation does no
better than Y alone, then the total variance equals the variance unaccounted for and
r2 = 0, indicating no ability to predict Yi (beyond the use of Y itself)
...
Smith and Gerald S
...
Even if b turns out to be zero, the predictions
are i = a = Y, which are the same as using the mean of criterion values in the first place
...
12 shows the residuals obtained after using the regression equation to
predict each value of Yt via its counterpart i
...
034 )2 + (-1
...
+ (0
...
09
Table 12
...
3

Rater
1
2
3
4
5
6
7
8
9
10
Mean

Predicted,*
4
...
464
3
...
377
3
...
034
6
...
148
7
...
377
4
...
491 + 0
...

This is the sum of squared errors in predicting Yi from
n

Residuals
Yi - i
-1
...
464
-1
...
377
2
...
034
1
...
148
1
...
623
0

i

i
...
3 )2 + (7 - 4
...
+ (2 - 4
...
10
This is the sum of squared errors in predicting Yi from Y
...
09
2
r yx1 = 1 76
...
723
and we say that 72% of the variation in Y has been accounted for by variation in X1
...
034 - 4
...
464 - 4
...
+ (1
...
3 )2

i=1

= 55
...

Figure 12
...
9 as well) put all these quantities in perspective by first
showing deviation of Yi - Y
...
10
...
09
...
01
...
09 + 55
...
10
...
Smith and Gerald S
...
10

Breakdown of Deviations Yi - Y into Two Additive Parts

Interpretation of Bivariate Regression
The sample data of Table 12
...
Table 12
...
The coefficients shown earlier
appear here, along with some other measures as well
...
13

Summary Output of Regression Analysis of Table 12
...

All requested variables entered
...

Dependent Variable: PREFEREN
Model Summary
Model
R
R Square

Method
Enter

Adjusted R Square

a

1
b

ANOVA
Model

a
...
850

...
688
df

Std
...
62354

Mean Square

1

Regression
55
...
087
8
Total
76
...

Predictors: (Constant), PROTEIN
b
...
Error
Model
1
(Constant)

...
979
PROTEIN

...
194
a
...
013
2
...

a

20
...
002

t


...



...
568

Overall, the linear regression equation is:
Y = 0
...
886 X1
The standardized coefficient for the independent variable (beta or ) is more
meaningful for multiple regression as it is a measure of the change in Y due to a unit
change in Xi when all independent variables have been “transformed” to the same units
...
This test was discussed in more detail in Chapter 11, in the
context of analysis of variance
...
Smith and Gerald S
...
630

...
If r2 is zero, then the sample r2 reflects
only sampling error and the F ratio will tend to be equal to 1
...
We first obtain the mean
squares, 55
...
636, by dividing the corresponding sums of squares by their
respective degrees of freedom
...
871
...
46 with an (illustrative) significance
level of 0
...
We reject the preceding null hypotheses and conclude that the overall
regression equation is statistically significant
...
There are marketing problems, however, where the dependent and
independent variables are rank orders or are best transformed into such rankings
...

We show the use of this measure by an example
...

Since the new method is easier to use, the manager wants to know if it will yield the same
relative results as the proven existing method
...
Table 12
...

Table 12
...
Applying this formula to our example, we get

319
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

If the subjects whose scores were used in computing rs were randomly drawn
from a population, we can test the significance of the obtained value
...
Under H0
any observed value would be due to chance
...
For our
example, we calculate

Looking at the table of critical values of t, we find that p
...
Thus, if a strict level is to be adhered to (
...

One final point concerning the use of the Spearman rank correlation coefficient is
warranted
...
When this happens, each of them is
applied (see Siegel, 1956, pp
...

The Spearman rank correlation coefficient is equivalent to the Pearson productmoment correlation coefficient with ranks substituted for the measurement observations,
X and Y
...

Finally, when the variables are nominally scaled, ordinal measures such as tau and
rs are not appropriate measures
...
One
association whose calculation and interpretation are straightforward
...
4)
...
We discuss lambda further in
the next section
...
Sometimes ordinal data are available and
as such are stronger than simple nominal measurement
...
Three regularly used tests are the Wilcoxon Rank Sum
(T), the Mann-Whitney U, and the Kolmogorov-Smirnov test
...

320
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The Wilcoxon T test is used for dependent samples in which the data are collected
in matched pairs
...
The Wilcox matchedpairs signed-ranks test gives more weight to pairs showing large differences between the
two measurements than to a pair showing a small difference
...
In addition, ordinal
measurement must hold for the differences between pairs
...
For instance, an
ordinal scaling device, such as a semantic differential, can be used to measure attitudes
toward, say, a bank
...
Changes in values of each scale could be analyzed by
this Wilcoxon test
...
This is a
relatively powerful nonparametric test, and is an alternative to the Student t test when the
analyst cannot meet the assumptions of the t test or when measurement is at best ordinal
...
As indicated earlier, results of U and t
tests often are similar, leading to the same conclusion
...

This test is sensitive to any kind of difference in the distributions from which the two
samples were drawn—differences in location (central tendency), dispersion, skewness,
and so on
...
Unfortunately, the
test does not by itself show what kind of difference exists
...
In this
case it is a goodness-of-fit test similar to single-classification chi-square analysis
...
Usually, however, we are interested in the strength of association as well as
the statistical significance of association
...
An association is substantively significant when it is
statistically significant and of sufficient strength
...
Although such judgment is subjective, it need not be
completely arbitrary
...
44)
...
The main descriptors for classifying the various indexes are
1
...
Whether one, both, or neither of the variables has categories that obey some
natural order (e
...
, age, income level, family size)
321
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

3
...
Rather, we shall illustrate one commonly
used index for 2 x 2 tables and two indexes that deal with different aspects of the larger R
x C (row-by-column) tables
...
Moreover, an added bonus is the
fact that phi equals the product-moment correlation—a cornerstone of multivariate
methods—that one would obtain if he or she correlated the two variables expressed in
coded 0 – 1 form
...
15, taken from a
study of shampoos
...
” We first note from the table that high frequencies appear in the cells: (a)
“body” included in ideal set and “no” to the question of whether her hair has enough
(natural) body; and (b) “body” excluded from the ideal set and “yes” to the same
question
...
15

Does Hair Have Enough Body Verses Body Inclusion in Ideal Set

Body included in ideal set
Body excluded from ideal set
Total

No
26 (A)
17 (C)

Hair Have Enough Body?
Yes
Total
8 (B)
34
33 (D)
50

43

41

84

Before computing the phi coefficient, first note the labels, A, B, C, and D
assigned to the four cells in Table 12
...
The phi coefficient is defined as

The value 0
...
Smith and Gerald S
...
However, in any given problem the upper limit of phi depends on the
relationships among the marginals
...
Looking at the letters (A, B, C, D) of Table
12
...
The more different the marginals, the lower the
upper limit that the (absolute) value of phi can assume
...
Indeed, the
absolute value of phi is related to chi-square by the expression

where n is the total frequency (sample size)
...
Note, however, that
phi, unlike chi-square, is not affected by total sample size because we have the divisor n
in the above formula to adjust for differences in sample size
...
This index is also related to chi-square and is defined as

where n is again the total sample size
...
15 we can first determine that chisquare is equal to 14
...
01
level
...

However, unlike the phi coefficient, the contingency can never attain a maximum value
of unity
...
707
...
87
...

323
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

However, like phi, the contingency coefficient is easy to compute from chisquare; moreover, like phi, its significance has already been tested in the course of
running the chi-square test
...

Occasions often arise in the analysis of R X C tables (or the special case of 2 2 tables)
where we desire to compute an asymmetric measure of the extent to which we can reduce
errors in predicting categories of one variable from knowledge of the categories of some
other variable
...

To illustrate the lambda-asymmetric coefficient, let us return to the crosstabulation of Table 12
...
Suppose that we wished to predict what category—no versus
yes—a randomly selected person would fall in when asked the question, “Does your hair
have enough body?” If we had no knowledge of the row variable (whether that person
included “body” in her ideal set or not), we would have only the column marginal
frequencies to rely on
...
As a consequence, we shall be wrong in
41 of the 84 cases, a probability error of 41/84 0
...
If we know that “body” is not included in the ideal set, we shall
predict “yes” and be wrong in 17 cases
...
We can consider this
error reduction relatively:

In other words, 39 percent of the errors in predicting the column variable are eliminated
by knowing the individual’s row variable
...

Lambda-asymmetric varies between zero, indicating no ability at all to eliminate
errors in predicting the column variable on the basis of the row variable, and 1, indicating
an ability to eliminate all errors in the column variable predictions, given knowledge of
the row variable
...
In
the case of Table 12
...
Smith and Gerald S
...

Finally, if desired, we could find a lambda-symmetric index via a weighted
averaging of C R and R C
...
Furthermore, lambdaasymmetric h is a natural and useful interpretation as the percentage of total prediction
errors that are eliminated in predicting one variable (e
...
, the column variable) from
another (e
...
, the row variable)
...
In the initial sections of this chapter, we introduced the basic concepts
of transforming raw data into data of quality
...

The focus of this discussion was heavily oriented toward how to read the data and how to
interpret the results
...

The first section of the chapter dealt with cross-tabulation and chi-square analysis
...
We next focused on the necessary statistical machinery to analyze
differences between groups: t-test, and one-factor and two-factor analysis of variance
...

We then looked at the process of analysis of variance
...
The
assumptions underlying various models were pointed out and a hypothetical data
experiment was analyzed to show how the ANOVA models operate
...
The concept of associations between two variables was introduced
through simple two-variable correlation
...
Several
alternative (but equivalent) mathematical expressions were presented and a correlation
coefficient was computed for a sample data set
...
Bivariate (two-variable) regression was discussed as the
foundation for the discussion of multivariate regression in the next chapter
...
Also, the Goodman and
Kruskal lambda measure for nominal measurement was briefly introduced, as were other
nonparametric analyses
...
Some of
these were shown in Table 12
...
Although somewhat dated, a useful reference for
selecting an appropriate statistical is the guide published by the Institute for Social
325
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Research at the University of Michigan (Andrews, et
...
, 1981 and its corresponding
software Statistical Consultant
...
78-80) presents a summary table of
which technique to use under which condition
...
M
...
, Davidson, T
...
, O’Malley, P
...
, & Rodgers, W
...
(1981)
...

Ann Arbor: Institute for Social Research, University of Michigan
...
F
...
Analyzing marketing research data with associated
models
...

Fink, A
...
How to manage, analyze, and interpret survey data
...

Gibbons, J
...
(1993)
...
Newbury Park, CA:
Sage
...
(1969, February)
...
American
Sociologist, 4, 44
...
A
...
H
...
Measures of association for cross
classification
...

Hellevik, O
...
Introduction to causal analysis: Exploring survey data
...

Lewis-Beck, M
...
(1995)
...
Thousand Oaks, CA: Sage
...
T
...
Use your brain when using a chi-square
...

Siegel, S
...
Nonparametric statistics for the behavioral sciences
...

Smith, S
...
(2005)
...
Thousand Oaks,
CA: Sage Publications
...
(1957)
...
New York: Harper & Row
...
Smith and Gerald S
...
The same
process applies to multivariate analysis, as is illustrated in Figure 13
...

Association implies only that two or more variables tend to change together to a greater
or lesser degree, depending upon the degree of association involved
...
It
should be understood, however, that association does not imply causation
...
Analytical techniques such as
causal-path analysis and structural equation models have been developed to aid in examining
possible causal relations in correlational data
...
d
...
Multivariate analyses arise when more than two variables are
to be analyzed at the same time
...
This use of multiple predictors allows the
researcher to conduct both multiple and partial regression
...
Like multiple regression, discriminant
analysis is a predictor model
...
In discriminant analysis, group membership (such as purchaser
and non-purchaser groups) is predicted using the multiple independent variables
...
We start with a broad overview of
multivariate procedures
...
For example, the
computation of the mean and standard deviation of a single variable is often done simply to
obtain a summary of the meaning of the entire set of values
...

In virtually all marketing research studies we are concerned with variation in some
characteristic, be it per capita consumption of soft drinks, TV viewing frequency, or customer
intention to repurchase
...
When
327
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

we analyze associative data, we hope to explain variation, achieving one or more of the
following:
1
...
Determine a function or formula by which we can estimate values of the criterion variable(s)
from values of the predictor variable(s)
3
...
We
may still be interested in the interdependence of a group of variables as a whole and the
possibility of summarizing information provided by this interdependence
...
1

Process for Selecting Appropriate Method of Analysis

A Classification of Techniques for Analyzing Associative Data
The field of associative data analysis is vast; hence, it seems useful to enumerate various
descriptors by which the field can be classified
...
1
...
We note that a data matrix is
often thought of as being in spreadsheet form (referred as a flat data matrix) and consists of a set
of objects (the n rows) and a set of measurements on those objects (the m columns)
...
The column variables are characteristics of
the objects
...
Cell values
may consist of nominal-, ordinal-, interval-, or ratio-scaled measurements, or various
combinations of these as we go across columns
...

Note however that some analysis software also has the ability to read relational databases that are
328
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

not in the traditional row x column spreadsheet format
...

Table 13
...


...

i

...


...


...

Xi1

...


...


...

Xi2

...


...


...

Xi3

...


...
j

...
X2j

...


...


...


...


...


...


...



...


...

Xim

...


...
Although not
exhaustive (or exclusive), the following represent the more common perspectives by which
analysis activities can be directed:
1
...
Focus of research: Is the emphasis on the objects (the whole profile or “bundle” of
variables), the variables, or both?
3
...

4
...
Type of association under study: Is the association linear, transformable to linear, or
nonlinear?
6
...
Whether the objects or variables of the data matrix are partitioned into subsets
or kept intact
2
...
Whether the variables are nominal-scaled or interval-scaled
4
...

Most decisions about associative data analysis are based on the researcher’s private
model of how the data are interrelated and what features are useful for study
...
Smith and Gerald S
...
) is predicated
upon prior knowledge of both the characteristics of the statistical universe from which the data
were obtained and the assumption structure and objectives associated with each statistical
technique
...
First,
concerning types of scales, all the multivariate techniques of this book require no stronger
measurement than interval scaling
...
While the original data may have been transformed by some nonlinear
transformation (e
...
, logarithmic), linear in the parameters means that all computed parameters
after the transformation are of the first degree
...


...
Note that b0 is a constant, while the other parameters, b1, b2,

...
Moreover, none of the bj’s depends on the value of either its
own Xj or any other Xk (k j)
...
1), to complex
...
This, in turn, can be broken down into
subcategories:
1
...


Single criterion/multiple predictor association
Multiple criterion/multiple predictor association

Multivariate techniques that deal with single criterion/multiple predictor association
include multiple regression, analysis of variance and covariance, two-group discriminant
analysis, and CHAID (chi-square automatic interaction detection)
...

Analysis of Interdependence
In some cases, we may not wish to partition the data matrix into criterion and predictor
subsets
...
Techniques
such as factor analysis are used if the focus of interest is on the variables of the (intact) data
matrix
...

Selecting a Technique

How does a researcher select the multivariate technique to use in a given situation? In
addition to the research question underlying the analysis, we suggested in Figure 13
...
Four major sets of factors influence the choice of an appropriate statistical
technique:
330
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Research parameters (analysis purpose and specifications)
Known characteristics of the data
Properties of the statistical technique
Researcher characteristics

Research parameters include the purpose of analysis and the kind of analysis required
...
This information is
matched with the properties of a particular statistical technique (robustness, assumptions
required, and design purpose)
...

Operationally, the researcher can be assisted by any of numerous printed or online
guides, such as that provided at http://www
...
com/wiki/index
...


MULTIPLE AND PARTIAL REGRESSION
Multiple regression extends the concepts of bivariate regression by simply adding
multiple predictor (independent) variables
...
Table 13
...

Table 13
...
1

Protein, X1

Vitamin D, X2

3
7
2
1
6
2
8
3
9
2
43
4
...
908

4
9
3
1
3
4
7
3
8
1
43
4
...
791

2
7
1
2
3
4
9
2
7
3
40
4
...
708

Multivariate Analysis Can Be Simple

Not all multivariate analyses need to involve
complex and sophisticated statistical techniques
...
In the B2B area, for
example, this could be done by asking customers
to rate a supplier’s performance on a set of
attributes and characteristics, and then asking
those customers how important each is to them
...


331
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The theoretical model now becomes
with parameters estimated by
All assumptions previously discussed for the bivariate regression model continue to hold
in the present case
...
This is because X1 itself will usually be
correlated with X2
...
Such will no longer be the case
...

However, now the plot appears in terms of a three-dimensional space
...
Figure 13
...
2
...
247 + 0
...
484 X2
Figure 13
...
3

Scatter Plot and Fitted Regression Plane

Single Correlation Between X1 (Protein) and X2 (Vitamin)

Protein
Protein Pearson Correlation
1
Sig
...

N
10
Vitamin Pearson Correlation

...
(2-tailed)

...
Correlation is significant at the 0
...


Vitamin

...
002
10
1

...
493 denotes the change in per unit change in X1 when X2
is held constant
...
484 denotes the change in per unit change in X2 when X1 is held
constant
...
493 0
...
This is because X1 and X2 are themselves correlated, and X1 is now forced to share
some of its Y-variable association with the second predictor, X2
...
Smith and Gerald S
...
493 and b2 0
...
247 is considerably more complicated than finding a single regression coefficient in bivariate
regression
...
2 shows the statistical analysis for a regression analysis
...

______________________________________________________________________________
EXHIBIT 13
...


Method

VITAMIN,
a
PROTEIN
All requested variables entered
...
885

R Square

...
721

Change Statistics
R Square
Change
F Change
Model
1

...
652
Predictors: (Constant), VITAMIN, PROTEIN

df1
2

df2
7

Std
...
53482

Sig
...
005

ANOVA
Model
1

Sum of Squares
Regression
59
...
490
Total
76
...
805
2
...
652

Sig
...
005

Coefficients
Unstandardized
Coefficients
Model
B
Std
...
247

...
493

...
484

...
Dependent Variable: PREFERENCE

Standardized
Coefficients
Beta

...
450

t
0
...
466
1
...


...
186

...
Exhibit 13
...
783
...
However, we
recall from Chapter 12 that if only X1 is employed, the (simple) coefficient of determination is 72
percent
...
Smith and Gerald S
...
838 (see Table 13
...

What about
the square root of
?

This is called the coefficient of multiple correlation: and is interpreted as the simple correlation
(i
...
, bivariate correlation) between Yi and where, as we know, the are the predicted values
obtained by the best linear composite of X1 and X2 in the least-squares sense; this composite is
given by the multiple regression equation
...
It may be
the case that a simpler model involving only X1 (or only X2) would be sufficient
...

The standard error is a measure of dispersion about the average partial regression
coefficient over repeated samplings of Y for a fixed set of values on each of the predictors
...
Conceptually, this statement means that the standard error of b1 increases as
X1 becomes more completely accounted for by the remaining predictor variables
...
466 in Exhibit 13
...
73/3
...
The test of significance of t1 is carried out by finding the tabular value of the
t-distribution (Table A
...
If we continue to use a
significance level of 0
...
415 and b1 is significant
...
397) is not significant at the 0
...
Note that in Exhibit 13
...
186
...
When considered alone, the t-test for protein is significant at the
...

The t-test for each individual partial regression coefficient tests whether the increment in
2
R produced by the predictor in question is significant when a model including the predictor (and
all other predictors) is compared with a model including all predictors but the one being tested
...
3 percent of the variation in the dependent variable
...
In fact, this is the case
...
723
...
06 (0
...
723)
...
2 shows the major output measures of interest in applied multiple
regression studies:
1
...
R2—both the sample-determined and the population-adjusted values
3
...
Smith and Gerald S
...
Individual t-tests and standard errors for testing each specific partial regression coefficient
5
...
The accounted-for variance contributed by each predictor, where any shared variance of each
predictor (beyond the first) is credited to the predictors that precede it, based on the researcher’s
order for including the predictor variables

In the sample problem R2was highly significant, but only the first variable X1 was needed
to account for most of the variance in Y
...
1
level
...
This has
ˆ
already been computed in Chapter 11 as Y = 0
...
886 X 1
...
However, decisions must still be made by
the user
...

The two most widely used approaches are all-variable regression (demonstrated above)
and stepwise regression, as described further in Exhibit 13
...
4
...
Similarly, the results of the regression analyses are often similar for a small
number of variables, becoming increasingly disparate as more variables are introduced
...
For
more detail contrasting the difference between the all-variable and stepwise regression
methodologies, see Green, Tull, and Albaum (1988, pp
...

______________________________________________________________________________
EXHIBIT 13
...

The theoretical and statistical shortcomings are inherent in the process and difficult to overcome (Curtis,
1990)
...
Limit the number of variables initially included in the predictor pool, eliminating some with
high intercorrelations
...
Use as large a sample as possible, ideally with 40–50 cases for each variable in the pool
...
Validate the model through a follow-up sample or by splitting the original sample
...
When multicollinearity is a problem, consider using factor analysis as a data reduction
technique
...
“Best” typically refers to maximizing the explained variance (R2) of the
dependent variable
...
It
starts building a model by selecting the variable most highly correlated with the dependent variable
...
As new variables are added, others may be dropped in a repetitive process until the bestfitting model is produced
...
Smith and Gerald S
...
First, from a substantive perspective, the best-fitting
model may not be the best—or even good
...
It would not be unusual to obtain three completely different bestfitting models from the
same data set using these methods
...

Another problem occurs when high levels of multicollinearity exist in the pool of independent
variables
...
The greater the multicollinearity, the more likely an
unreasonable and unstable best-fitting model is selected
...
The problem is
that stepwise methods are purely data-driven, after-the-fact techniques not based on a prior theoretical
model
...


______________________________________________________________________________

______________________________________________________________________________
Summary Output of Stepwise Regression Analysis

EXHIBIT 13
...
2

Variables Entered/Removed
Model
1

Variables
Entered
VAR00002

Variables
Removed

...
050, Probability of F to remove>=
...


Dependent Variable:VAR00001

Model Summary
Model
1

R
a

...
723

Adjusted R Square

...
723
20
...
Error of the Estimate
1
...
F Change

...
013
Residual
21
...
100
Predictors: (Constant), VAR00002
Dependent Variable: VAR00001

df
1
8
9

Mean Square
55
...
636

F
20
...

a

...
Error
1
(Constant)

...
979
VAR00002

...
194
Dependent Variable: VAR00001

Standardized
Coefficients
Beta

...
501
4
...


...
002

Excluded Variables
Model
Beta In
t
a
1
VAR00003

...
397
Predictors in the Model: (Constant), VAR00002
Dependent Variable: VAR00001

Sig
...
205

Partial
Correlation

...
298

______________________________________________________________________________

336
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Multicollinearity
Put simply, multicollinearity refers to the problem in applied regression studies in which
the predictor variables exhibit very high correlation among themselves
...

Unless one is dealing with experimental design data, it is almost always the case that
predictor variables in multiple regression will be correlated to some degree
...

The study of multicollinearity in data analysis revolves around two major problems: (a)
How can it be detected; and (b) what can be done about it? These problems are particularly
relevant to marketing research, where one often faces the dilemma of needing a large number of
variables to achieve accuracy of predictors, and yet finds that as more predictors are added to the
model, their intercorrelations become larger
...
Some
researchers have adopted various rules-of-thumb: For example, any pair of predictor variables
must not correlate more than 0
...
While looking at
simple correlations between pairs of predictors has merit, it can miss more subtle relationships
involving three or more predictors
...
Usually one
would want to guard against having any of these multiple correlations exceed the multiple
correlation of the criterion variable with the predictor set
...
Ignore it
...
Delete one or more of the offending predictors
...
Transform the set of predictor variables into a new set of predictor-variable
combinations that are mutually uncorrelated
...
First, one can have
multicollinearity in the predictor variables and still have strong enough effects that the estimating
coefficients remain reasonably stable
...
A prudent procedure in checking one’s predictor set for multicollinearity is to examine
the standard errors of the regression coefficients (which will tend to be large in the case of high
multicollinearity)
...
Third, most computer regression routines incorporate checks for serious
multicollinearity (see SPSS Linear Regression Analysis Statistics Help for a list of diagnostic
analyses); if the program does not indicate this condition, the researcher can generally assume
that the problem is not acute
...
Usually, because of their high intercorrelations with
the retained predictors, the overall fit will not change markedly
...
Smith and Gerald S
...

Methods also exist for transforming the original set of predictors to a mutually
uncorrelated set of linear composites—for example, principal components analysis (which is
discussed in Chapter 14)
...
If all components are retained, the predictive accuracy will be precisely the same as
that obtained from the original set of predictors
...
Another possibility, of course, is to use
only one of the variables for each component to represent that component in the regression
analysis
...
We have commented in this
chapter on the tendency of regression models (and the same is true of other multivariate
techniques) to capitalize on chance variation in the sample data
...

Cross-validation is a simple procedure for examining whether the regression equation
holds up beyond the data on which its parameters are based
...
The regression equation is then computed
from the remaining data
...
The researcher then finds the simple coefficient of determination between the
Yi in the hold-out sample and their predicted Yi counterparts
...
Doing this requires a
larger sample size
...
Split the cases randomly into halves
...
Compute separate regression equations for each half
...
Use the first-half equation to predict the second-half Yi values
...
Use the second-half equation to predict the first-half Yi values
...
Examine each partial regression coefficient across split halves to see if agreement is
obtained in both directions (algebraic sign) and in magnitude
...
Compute a regression equation for the entire sample, using only those variables that
show stability in the preceding step
...

Depending on the sample size, of course, one could split the sample into thirds, quarters,
and so on
...
Even so, single and double crossvalidation are extremely useful undertakings, and, with a few clicks of a mouse, are easy to
implement
...
Smith and Gerald S
...
Their results indicated that
splitting the data into halves is suboptimal
...


TWO-GROUP DISCRIMINANT ANALYSIS
When dealing with associative data, the marketing researcher may encounter cases where
the criterion variable is categorical, but where the predictor variables involve interval-scaled
data
...
Other potential applications
also come to mind:
Two Groups:
How do consumers who are heavy users of my brand differ in their demographic profiles
from those who are nonusers?
How do respondents who show high interest in a new set of concept descriptions differ in
their readership levels of certain magazines from those who show low interest?
How do homeowners who select a variable rate mortgage differ in their demographic profiles,
mortgage shopping behavior and attitudes, and preferences for mortgage features from
homeowners selecting a conventional fixed-rate mortgage?
Multiple Groups:
Are significant demographic differences observed among purchasers of Sears, Goodyear,
Goodrich, and Michelin tires?
How do doctors, lawyers, and bankers differ in terms of their preference ratings of eight
different luxury automobiles?
How can loyal shoppers of Nordstrom’s, Macy’s, and Neiman-Marcus be distinguished on
the basis of their attitudes about each retailer?

Many other such problems could be added to the list
...
It is also assumed we know that person’s profile on a set of (assumed or otherwise)
interval-scaled predictor variables, such as age, income, years of education, or other background
variables
...
Here we shall assume that the function is linear
...

Objectives of Two-Group Discriminant Analysis
Two-group discriminant analysis (and classification) involves four main objectives:
1
...
Establishing procedures for assigning new individuals, whose profiles (but not group identity)
are known, to one of the two groups
339
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

3
...
Determining which variables account most for intergroup differences in mean profiles

These objectives are the bases for very distinct, purposes and procedures for conducting
discriminant analysis (Smith, 2004)
...

The second procedure, discriminant classification analysis, uses the predictive functions
derived in the first procedure to either classify fresh sets of data of known group membership,
thereby validating the predictive function, or if the function has previously been validated, to
classify new sets of observations of unknown group membership
...
4
...
4

Stages of Discriminant Analysis

Stages

Purpose
Assumptions of linear
Discriminant model
...
No
classifies
classification involved (biased)

Classification
analysis of
new data
set of known
groupings
(1) Classify data using
classification rule derived
from predictive function
(2) May be part of
validation analysis of
initial predictive function

No validation
required

Validation required

No validation
required

Classification
analysis of
new data
set of unknown
groupings
(1) Classify data using
classification rule derived
from predictive function
(2) May be part of
validation analysis of initial
predictive function
Initial predictive function
must have been previously
validated

Geometric Representation
If we have n persons measured on m variables, the profile of each person can be
portrayed as a point in m dimensions
...
The less overlap noted among
intergroup profiles in that space, the more likely it is that discriminant analysis can help us
separate the groups
...
4 provides one way to show what
happens when a two-group discriminant function is computed
...
We could plot in the
scatter diagram the association of variable X2 with X1 for each group, maintaining group identity
by the use of filled-in dots or open circles
...
If a straight line is drawn through the two
points where the ellipses intersect and is then projected to a new axis Z, we can say that the
overlap between the univariate distributions A´ and B´ (represented by the shaded area) is
smaller than would be obtained by any other line drawn through the ellipses representing the
scatter plots
...

340
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 13
...
4 is that the Z-axis expresses the twovariable profiles of groups A and B as single numbers
...
Thus, the axis Z
condenses the information about group separability (shown in the bivariate plot) into a set of
points on a single axis
...

In most problems, we have more than two predictor variables
...
In any case, the basic objective is still to find one axis in the mdimensional space that maximally separates the centroids of the two groups after the points are
projected onto this new axis
...
Analogously, in twogroup discriminant analysis we try to find a linear composite of the original variables that
maximizes the ratio of among-to-within groups variability
...

A Numerical Example

Let us return to the example involving ready-to-eat cereals that was first presented in
Table 13
...
As recalled, we wished to see if amount of
protein and vitamin D influenced consumers’ evaluations of the cereals
...
The (hypothetical) data appear in
Table 13
...
If we were forced to choose just one of the axes, it
would seem that X1 is a better bet than X2
...
Smith and Gerald S
...

Table 13
...
4

Like
Like
Like
Like
Like

7
8
9
10
11
9

6
4
7
6
9
6
...
5
3
...
4
2
...
5 shows a scatter plot of the X1 and X2 data of Table 13
...
We note that perfect
discrimination can be achieved with X1 if we erect a line perpendicular to the horizontal axis
between the scale values of 6 and 7
...
Given this picture, we would not be surprised if the best
linear composite places a considerably larger weight on X1 than on X2
...
5

Scatter Plot of Two-Group Sample Data of Table

Why not use X1 alone, rather than a composite of X1 and X2? First, the data of Table 13
...
Second, we have not
explicitly taken into consideration either the variability of X1 versus X2 or their correlation
...
Smith and Gerald S
...

As noted earlier, the key problem of two-group discriminant analysis is to find a new axis
so that projections of the points onto that axis exhibit the property of maximizing the separation
between group means relative to their within-groups variability on the composite
...

Using a discriminant analysis program, we solve for the discriminant weights d1 and d2
that maximize the separation between the groups
...
The general form for a linear discriminant function is Z = d0 +
d1X1 + d2X2, where d0 is the intercept or constant term in the function
...

Having found the discriminant function, it is a straightforward procedure to find discriminant
scores for the centroids of the two groups and the grand mean:

We may also apply this procedure to each of the 10 pairs of X1, X2 values in Table 13
...
For example, the discriminant score for the first case in the disliker group is
Z = 0
...
147(4) = 0
...
6
...
368, – 0
...
e
...

Alternatively, one could use any point with coordinates proportional to (0
...
147)
...
6 differs
from the original unit in which X1 and X2 are expressed
...
The original points can then be projected onto this new axis
...
(Similarly, all ten original points could be projected onto the
343
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

discriminant axis as well
...
5 times the (absolute-value) weight (d1 0
...
147 to X1) as is
given to X2
...
6

Plot of the Discriminant Axis and Point Projections

The analysis output is shown in Exhibit 13
...
Note that the discriminant function
coefficients are different from those shown earlier, meaning that the discriminant scores for each
person also will differ
...
The unstandardized discriminant function coefficients evaluated
at group means sum to zero
...

EXHIBIT 13
...
Deviation
GRP 1 PROTEIN
4
...
58114
VITAMIN
4
...
81659
GRP 2 PROTEIN
9
...
58114
VITAMIN
6
...
81759
Total
PROTEIN
6
...
02765
VITAMIN
5
...
01108
Tests of Equality of Group Means
Wilks’
Lambda
F
df1
df2
PROTEIN

...
000
1
8
VITAMIN

...
030
1
8
Summary of Canonical Discriminant Functions
Eigenvalues
Function
Eigenvalue
% of Variance
a
1
3
...
0
First 1 canonical discriminant functions were used in the analysis
...
206
11
...
000
5
5
...
000
5
5
...
000
10
10
...


...
120
Cumulative %
100
...
891

Sig
...
004

344
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Standardized Canonical Discriminant Function Coefficients
Function
1
PROTEIN
1
...
608
Structure Matrix
Function
1
PROTEIN

...
313
Pooled within-groups correlations between discriminating variables and standardized canonical discriminant
functions
...

Canonical Discriminant Function Coefficients
Function
1
PROTEIN

...
335
(Constant)
–3
...
00
–1
...
00
1
...
00

...
000
2
...
500
5
5
...
000
10
10
...
00
2
...
035
3
...
706
–
...
317
–17
...
00
2
...
00
5
0
5
2
...
00
100
...
0
100
...
00

...
0
100
...
00
5
0
5
2
...
00
100
...
0
100
...
00

...
0
100
...
In cross validation, each case is classified by the
functions derived from all case other than that case
...
0% of original grouped cases correctly classified
...
0% of cross-validated grouped cases correctly classified
...
Smith and Gerald S
...
The classification problem, in turn, involves two
additional questions: (a) how well does the function assign the known cases in the sample; and
(b) how well does it assign new cases not used in computing the function in the first place?
To answer these questions we need an assignment rule
...
6
...
To
classify an individual, if Zi Zcrit then the individual belongs in one group, while if Zi Zcrit then
the individual goes into the other group
...
6, no misassignments
will be made if we adopt the rule:
Assign all cases with discriminant scores that are on the left of the midpoint (1
...

Assign all cases with discriminant scores that are on the right of the midpoint (1
...

That is, all true dislikers will be correctly classified as such, as will all true likers
...
6
...
For example, had any of the five
true dislikers been called likers, the first row and second column would contain not a zero but,
rather, the number of such misassignments
...
824 for dislikers and 2
...

For any new case, compute the discriminant score and assign the case to that group
whose group score is closer
...

If a higher probability existed for likers, we could reduce the expected probability of
misclassification by moving the cutting point of 1
...
824, the mean score
for dislikers) so as to give a wider interval for the larger (likers) group
...
824
...
e
...
That is, some capitalization
on chance takes place in discriminant analysis, and one needs a way to measure this bias of an
inflated percentage of correctly classified observations, just as cross-validation should be
employed in multiple regression
...
Smith and Gerald S
...

The most frequently suggested validation approach is the holdout method, in which the
data set is randomly split into two subsamples
...
It is
suggested that this split-sample validation be replicated by using a different convention for the
random assignment of the observations
...
These problems relate to the number of variables included in
the discriminant function (when a stepwise procedure is used) and to inferences that the
researcher might make concerning the classificatory power of resulting discriminant functions
...
Also,
there can be a wide range in the correct classification performance of the validation replications,
although the analysis functions are quite stable
...
For the analysis
and validation samples there is a measure of the percentage of cases that would have been
correctly classified based on chance alone
...

Because our “Cereal Data” achieved 100 percent correct classification, we will now
switch to another example to demonstrate these measures of chance classification
...
6 was derived from a
discriminant analysis where accident insurance purchasers and non-purchasers were predicted
using demographic variables
...
16 percent of all subjects surveyed were correctly classified as purchasers and
non-purchasers of accident insurance
...
001 level (
24
...
58, df 1)
...

Tests of Group Differences
Morrison (1969) considered the question of how well variables discriminate by
formulating a likelihood ratio to estimate chance classification
...
This proportion, designated the proportional chance criteria,
or Cpro (Morrison, 1969), is expressed as where
the proportion of customers in the sample categorized as purchasers
p the true proportion of purchasers in the sample
(1
) the proportion of the sample classified as non-purchasers
(1 p) the true proportion of non-purchasers in the sample
This likelihood analysis states that 65
...
The proportional chance criterion, Cpro, has been
used mainly as a point of reference for subjective evaluation (Morrison, 1969), rather than the
basis of a statistical test to determine if the expected proportion differs from the observed
proportion that is correctly classified
...
Smith and Gerald S
...

Thus for the example problem, the difference between expected and actual overall correct
classification is significantly different at the
...
This overall test of significance suggests
that further analysis should be conducted to determine the source of the deviation from chance
expectations
...
6

Classification Table for Accident Insurance Purchasers

Fequency, Row %
Chi-Square Contribution
Actual
Purchase

Predicted Purchase

Predicted
Non-Purchase
n=22
n=11
66
...
3%
15
...
46
Actual
N=66
N=199
Non-Purchase
24
...
1%
1
...
80
Column Totals
n=88
n=21
Column Percentage
29
...
5%
Percent of Cases Correctly Classified = 221/298 = 74
...
599 df=1, Significance <
...
1%
n=265
88
...
e
...

Classification and Misclassification Within Groups
The analysis to determine the source of deviation is conducted using the maximum
chance criterion, designated Cmax (Morrison, 1969)
...
The computation of Cmax is based on the
assumption that all observations are categorized as coming from that group: for example, given
that all 298 purchasers and non-purchasers were classified as purchasers, then the maximum
correct classification, Cmax, would be expressed

Because we are interested in the correct classification of insurance purchasers, the test of
classification involves asking if the 66
...
1-percent maximum expected chance classification
...


348
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

* Significant at the
...

This test may also be conducted for the other cells in the classification matrix:

Thus cell Z11 shows that observed classification is significantly greater than is expected
to occur by chance classification alone
...
Thus the discriminant functions appear to shift the classification
of subjects toward the purchaser categories, as demonstrated by significantly greater than
expected classification in the upper and left portions of the classification matrix
...
These coefficients are the basis of classifying, profiling, and
evaluating the underlying discriminant dimensions
...
The jackknife statistic and the U-method have been
proposed as such procedures
...

When a discriminant analysis is run using a computer program, there is another option for
validation, especially when a small sample is involved
...
While we do not deal specifically
with the classification of cases other than those from which the discriminant function is
developed, it should be recognized that once the discriminant function is developed and validated
for the population it may be used to classify other groups within the population
...
This is analogous to testing for the significance of R2 in multiple regression
...
We do not delve
into the technical details of Mahalanobis squared distance, other than to say that it is like
ordinary (Euclidian) squared distance that is computed between two centroids in a space with
correlated axes and different measurement units
...
The canonical correlation coefficient is a measure of the association that
349
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

summarizes how related the discriminant function is to the groups
...
An indirect, and most widely used, approach to test for
the statistical significance of the discriminant function examines the ability of the variables to
discriminate among the group beyond the information that has been extracted by the previously
computed functions
...
Wilks’ lambda is a multivariate measure of group
differences over discriminating variables and can be calculated in several ways
...
0 the group centroids are equal and there is no discrimination (Klecka,
1980)
...
Moreover, tests of statistical significance are sensitive to sample size
...
Finally, there is a need to go beyond statistical significance and
test for the practical (or substantive) significance
...

Relative Importance of Predictor Variables
Because the original variables X1 and X2 were expressed in different units and display
different standard deviations as well, the analyst generally standardizes the discriminant weights
before examining their relative importance
...

These coefficients are not appropriate to assess the relative discriminatory power of the variables
included in the analysis
...
Since a relative
importance value shows the value of a particular variable relative to the sum of the importance
values of all variables, the relative importance of a variable (Rj) is given by Aw and Waters
(1974):
350
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

The end result of using this procedure is shown in Table 13
...
As
these data show, although nine variables were included in the discriminant function, slightly
more than 88 percent of the total discrimination was accounted for by only three variables
...
7 Relative Importance of Consumption Values for Teenage Smokers and Nonsmokers
Variables
(Factors)
1
2
3
4
5
6
7
8
9

Standardized
Coefficients
-0
...
091
0
...
344
0
...
018
-0
...
007
0
...
299
0
...
864
0
...
366
-0
...
007
0
...
0989
0
...
2861
0
...
1211
-0
...
0014
-0
...
0533

Never
Smoked
Mean (X2)
0
...
0277
-0
...
1050
-0
...
0056
0
...
0021
-0
...
0567
0
...
4741
0
...
0850
0
...
0000
0
...
0165
0
...
9%
0
...
4%
10
...
9%
0
...
0%
0
...
3%
100
...
2% (cross-validated grouped cases); Cpro – 51
...
886 (prob
...
337
Source: Albaum, Baker, Hozier and Rogers, 2002, p
...


Determining relative importance of the predictor variables in discriminant analysis
becomes increasingly difficult when more than two groups are involved
...

Multiple Discriminant Analysis
All of the preceding discussion regarding objectives and assumption structure applies to
multiple discriminant analysis as well
...
What primarily distinguishes multiple discriminant analysis from the twogroup case is that more than one discriminant function may be computed
...
In general, with G groups and m predictors we
can find up to the lesser of G 1, or m, discriminant functions
...
Moreover, it
turns out to be a characteristic of multiple discriminant analysis that the first function accounts
for the highest proportion of the among-to within-groups variability; the second function, the
next highest; and so on
...
There remains the
problem of interpretation of the functions
...
Smith and Gerald S
...
Interested readers should consult more extensive and advanced discussions
of the topic such as Johnson and Wichen (2002), Hair, Tatham, Anderson, and Black (1998), and
Stevens (1996)
...
This tool
allows the researcher to explore relationships where both the number of respondents and the
number of variables is very large
...
When interaction
effects exist, the simple additive property of individual predictor-variable contributions to
changes in the criterion no longer holds
...
CHAID is insensitive to various forms of interaction
...

While the predictor variables might originally be (a) nominal-, (b) ordinal-, or (c) interval
scaled, these predictors are all recorded into categorical (nominal-scaled) variables with one of
two characteristics:
1
...
Order across classes can be retained
For example, if one of the original predictors is employment status (white-collar, bluecollar, unemployed) one may wish to treat this variable as an unordered polytomy
...
In this case
one would probably wish to maintain order across classes
...
It has been
suggested that the primary goals of CHAID are segmentation analysis and an exploration of the
relationships between variables
...

There are more than just a few potential explanatory variables
...

AID also offers a variety of uses:
Identification of demographic variables that discriminate and predict the dependent variable
category, such as heavy users
Identification of interaction effects for further analysis, such as logistic regression
Data mining to find significance among the hundreds of possible crosstabulation
combinations
Identification of predictor variables that contribute little to explaining the dependent variable
352
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Basically, CHAID performs a series of contingency table analyses of the data (similar in
spirit to the bivariate cross-tabulation analyses discussed in Chapter 12)
...
At each step of the sequential splitting and merging process, CHAID
looks for the best available split, not the best set of some number of final groups
...
The predictor variable leading to the best binary split and how that predictor is split
2
...
The criterion-variable splits on each of the two subgroups
An Example of CHAID

Advertising executives have conducted research to identify profitable segments for a
sweepstakes promotion
...
15 for nonrespondents for mailing costs
...
That is, when asked
to consider the subscription to magazines through the associated sweepstakes plan, which market
segments are most likely to subscribe?
A large database (n 81,040) was available that contained information about their
response to the sweepstakes offer and known information about eight different demographic and
credit card ownership predictors
...
”
CHAID was applied to this large data bank, and the results are shown in the tree diagram
of Figure 13
...
At the top of the diagram we note that the total-sample probability of being a
subscriber is 0
...
That is, 478 out of the total of 81,040 recipients of the sweepstakes offer
paid for a subscription
...

Figure 13
...

These splits are detailed in Table 13
...

We observe that Node 8 provides the most gain (profitability) and is defined by the
predictor variables as a two-person household, headed by an individual aged 55–64 with income
of $15,000 or more
...


353
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 13
...
Smith and Gerald S
...
8

CHAID Tree Summary Map

Table 13
...
76
0
...
47
1
...
39
2
...
82
31
...
72
0
...
69
0
...
19
0
...
06
0
...
03
-0
...
15

3628
...
90367
2381
...
12275
554
...
87554
271
...
00454
-283
...
99468

619
1269
2460
3942
10742
13101
15384
40768
79938
80209

0
...
57
3
...
86
13
...
17
18
...
31
98
...
97

0
...
59
0
...
40
0
...
19
0
...
08
0
...
02

Index (%)
3628
...
81316
2745
...
45705
1118
...
02898
868
...
51311
115
...
22261

Node 12 is a 3–4 person household with a young to middle-age head of household who
holds a bankcard
...
We conclude that we should not
target the large family segment having five or more family members (Node 18), or the poor
empty-nester segment that is identical to our most profitable segment (Node 8), except without
the income base
...
31
percent of our database that promises to be profitable and another 49
...
We further observe that the average gain (profit) drops
from 0
...
08 with the inclusion of Node 1, which has an average gain of 0
...
The successive targeting of the next most profitable segment can do much to
maximize profits, as can eliminating the negative gain segments
...

In short, this kind of information underscores the importance of certain types of
demographic variables in this case, and leads to a more detailed understanding of the kinds of
segments that responded by subscribing to this sweepstakes magazine offer
...
g
...
It should be emphasized that a basic assumption
underlying the application of CHAID is that variables useful in explaining one part of the
database are not necessarily those most effective for explaining another part
...
Smith and Gerald S
...
All partitionings of the sample may be subject to a requirement that the proportionate
reduction in the criterion-variable probability exceed some level (specified by the
researcher)
...

2
...
This is to guard against splits that,
pragmatically speaking, are not worth the effort (for example, where the group is
already quite homogeneous)
...
In addition to the control parameters above, the researcher may place an upper limit
on the total number of groups formed and/or the minimum number of persons (or
objects) in each group
...

First, CHAID is generally designed for really large samples, on the order of 1,000 or more
...
Moreover, no statistical inferential apparatus is
associated with the approach
...

Second, CHAID, being a sequential search procedure, does not specify an explicit model
in the way, for example, that Discriminant Analysis does
...
This can then be followed by the formulation of an explicit
logit regression model that includes main effect and interaction terms of specific interest to the
research
...

Third, despite the explicit concern for interaction, CHAID is found to be insensitive to
various forms of interaction
...

One of the nice characteristics of CHAID is its simplicity of output, which facilitates an
ease of understanding by researchers and managers alike
...
This is illustrated in
the brief case study analyzed using the SPSS program AnswerTree, a suite of programs that
includes CHAID and CART models
...
0 program
...


356
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Canonical Correlation Analysis
Canonical correlation is a generalization of multiple correlation analysis to two or more
criterion (dependent) variables
...

What we would like to find out is how highly correlated the group of two criterion
variables is with the group of three predictors
...

This is what canonical correlation is all about
...

Thompson (1984, p
...
9 shows the results of the radial tire problem
...
Thus, in our case we would obtain two pairs of linear composites,
uncorrelated across pairs, with the first pair exhibiting maximum correlation
...
All composites are
mutually uncorrelated across pairs
...
05 alpha level; hence, only this pair of weights is
shown
...
Smith and Gerald S
...
9

Result of Canonical Correlation
Canonical Weights

Criterion set
Y1
Y2
Predictor set
X1
X2
X3
Canonical correlation
Input Correlations For Canonical Analysis
Y1
Y2
Y1
1
...
659
1
...
202
0
...
097
0
...
321
0
...
110
1
...
594
0
...
346
0
...
817
0
...
539
0
...
930

X1

X2

X3

1
...
086
0
...
000
0
...
000

Table 13
...
582
...
Since mutual association between the pair of batteries is involved, we can
say that the pair of linear composites account for (0
...

The canonical weights for the criterion set show that Y2 (post-exposure interest) is the
dominant variable in the criterion set; its canonical weight is 1
...
The dominant variable in the
predictor set is X3; its weight is 0
...
Since all variables are standardized to zero mean and unit
standard deviation before the analysis, the weights are already in standardized form
...
9 really says that if we formed a linear composite of the criterion variables using
the canonical weights:
0
...
069Ys2
Tc
and another linear composite of the predictors, using the canonical weights:
Tp 0
...
141Xs2 0
...
582 (the canonical correlation)
...
9
...
Again
we note that Y2 is the most important variable (structure correlation of 0
...
930) in the predictor set
...
539, almost as
high as the correlation between the full batteries
...

Probit and Logit Analysis
At the beginning of this chapter we emphasized that multiple regression is a useful technique for
estimating a dependent variable from a set of independent variables
...
When the dependent variable is dichotomous, binary regression can be used
...
Smith and Gerald S
...
e
...
e
...
To overcome these
problems, probit and logit have been developed
...
They differ solely in the
assumption made about the frequency distribution of the response:
In probit, the response is assumed to be normally distributed
...


Reported research about marketing applications seems to indicate that logit is the favored
technique of the two (Malhotra, 1984)
...

Explanations of how these techniques work and the underlying theories behind them tend
to be technical
...

Path Analysis/Causal Modeling
In recent years, marketing researchers have become increasingly interested in the useful
application of path analysis and structural equation modeling as an approach to causal modeling
...
From all this interest has emerged the development of structural
equation modeling procedures
...

The essence of what happens with this type of analysis can be seen from a simple
illustration
...
10 we show a system with four unobservable constructs, each of which
has more than one observable measure
...

To overcome this problem and better estimate the relevant parameters, some rather
complicated procedures have been developed which estimate both between the constructs (bij)
and their measures and the links between the constructs themselves (aij) simultaneously
...


359
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 13
...
, Baker, K
...
, and Rogers, R
...
Copyright © 2002
...


Another technique is PLS (partial least squares)
...
These techniques, and structural
equation modeling in general, are highly technical methods of analyses and are recommended
only for experienced users
...
Our discussions of
multiple regression and discriminant analysis have been more detailed than the descriptions of
other techniques
...

We started the chapter with a discussion of multiple regression and described such
measures as regression coefficients, the coefficient of determination, and the product-moment
coefficient
...
We described a sample problem and solved it numerically
...

This chapter concluded with a brief introduction to some less well-known techniques for
analyzing between-set dependence: CHAID, canonical analysis, correspondence analysis, and
probit and logit analysis
...

We have tried to discuss the principal assumption structure of each technique, appropriate
problems for applying it, and sufficient numerical applications to give the reader a feel for the
kind of output generated by each program
...

Our coverage of so vast and complex a set of methods is limited in depth as well as
breadth
...

This grounding will probably embrace three facets: (a) theoretical understanding of the
techniques; (b) knowledge of the details of appropriate computer algorithms to implement the
techniques; and (c) a grasp of the characteristics of substantive problems in marketing that are
relevant for each of the methods
...

360
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

REFERENCES
Aaker, D
...
(1981)
...
In Multivariate analysis in marketing
(2nd ed
...

Albaum, G
...
G
...
C
...
, & Rogers, R
...
(2002)
...
Journal of Consumer Affairs, 36(1), 50–76
...
H
...
D
...
Linear probability, logit, and probit models
...

Aw, R
...
, & Waters, D
...
A discriminant analysis of economic, demographic, and
attitudinal characteristics of bank charge-card customers
...

Bagozzi, R
...
Causal models in marketing
...

Berk, K
...
, & Carey, P
...
Data analysis with Microsoft Excel
...

Blalock, H
...
, Jr
...
Causal inferences in nonexperimental research
...

Carroll, J
...
, Green, P
...
, & Schaffer, C
...
(1987)
...
Journal of Marketing Research, 24, 445–450
...
D
...
E
...
M
...
Interpoint distance comparisons in
correspondence analysis
...

Clausen, S
...
(1998)
...
Thousand Oaks, CA:
Sage
...
R
...
D
...
(1979, February)
...
Journal of Marketing Research, 14, 60–68
...
E
...
3)
...
Marketing News, 24, 26
...
E
...
F
...
G
...
Bias in multiple discriminant
analysis
...

Green, P
...
, Tull, D
...
, & Albaum, G
...
Research for marketing decisions (5th ed
...

Greenacre, M
...
(1993)
...
London: Academic Press
...
F
...
L
...
E
...
(1998)
...
Upper Saddle River, NJ: Prentice Hall
...
Smith and Gerald S
...
A
...
W
...
Applied multivariate statistical analysis (5th ed
...

Klecka,W
...
(1980)
...
Beverly Hills, CA: Sage
...
A
...
Discriminant analysis
...

Levine, M
...
(1977)
...
Beverly Hills, CA: Sage
...
F
...
Interpreting probability models: Logit, probit, and other generalized linear
models
...

Magidson, J
...
(1993)
...
0
...

Malhotra, N
...
(1984, February)
...
Journal of
Marketing
Research, 21, 20–31
...
(2001)
...

Thousand Oaks, CA: Sage
...
B
...
M
...
d
...
Working
Paper, Virginia Polytechnic Institute and State University, Blacksburg, VA
...
G
...
On the interpretation of discriminant analysis
...

Mosteller, F
...
L
...
Influence in an authorship problem
...

Smith, S
...
(2004)
...
Retrieved May 18, 2004, from
http://marketing
...
edu/htmlpages/ tutorials/discriminant
...
H
...
R
...
Crossvalidating regression models in marketing
research
...
) Stern School of Business, New York University
...
(1996)
...
Mahwah, NJ: Lawrence
Erlbaum
...
(1984)
...
Beverly Hills,
CA: Sage
...
, & Smith, S
...
Using correspondence analysis to map relationships:
It’s time to think strategically about positioning and image data
...

Whitlark, D
...
(2004)
...
Provo, UT: Marriott School, Brigham Young Univ
...
Smith and Gerald S
...
Rather, it is the entire set of interdependent relationships that
are of interest
...
This methodology is useful to identify consumer lifestyle and
personality types
...
The fundamentals of cluster
analysis are described using examples of respondents and objects grouped because they have
similar variable scores
...
Multidimensional scaling methodology is closely related to
factor analysis, while conjoint analysis uses a variety of techniques (including analysis of
variance designs and regression analysis) to estimate parameters, and both techniques are related
to psychological scaling (discussed in Chapter 9)
...


AN INTRODUCTION TO THE BASIC CONCEPTS OF FACTOR ANALYSIS
Factor analysis is a generic name given to a class of techniques whose purpose often
consists of data reduction and summarization
...

Factor analysis operates on the data matrix
...
The most widely
used mode of factor analysis is the R-technique (relationships among items or variables are
examined), followed distantly by the Q-technique (persons or observations are examined)
...
1
...

The P- and O-techniques might be appropriate for looking at the life cycle of a product class, or
perhaps even changes in demographic characteristics of identified market segments
...
1

Modes of Factor Analysis

Six distinct modes of factor analysis have been identified (Stewart, 1981, p
...
The alternative modes of
factor analysis can be portrayed graphically
...
R-type and Q-type techniques deal with the variables/persons dichotomy (b)
...


363
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Factor analysis does not entail making predictions using criterion and predictor variables;
rather, interest is centered on summarizing the relationships involving a whole set of variables
...
The analyst is interested in examining the strength of the overall association among variables,
in the sense that a smaller set of factors (linear composites of the original variable) may be
able to preserve most of the information in the full data set
...

2
...

3
...


The major substantive purpose of factor analysis is to search for (and sometimes test)
structure in the form of constructs, or dimensions, assumed to underlie the measured variables
...
Common
factors are the underlying structure that contributes to explaining two or more variables
...
This non-common factor variance is called a unique
factor and is specific to an individual variable
...
1, in which four variables are reduced to two factors that summarize the majority of
the underlying structure, and four unique factors containing information unique to each variable
alone
...
Smith and Gerald S
...
In some applications the researcher may find that the variables are so highly
correlated that a single factor results
...
In response to these eventualities, the
researcher may revise the variable list and add additional variables to the factor analysis
...
Behavioral and consumer researchers have employed these methods
to develop measurement instruments such as personality profiles, lifestyle indexes, or measures
of consumer shopping involvement
...

We use a numerical example to illustrate the basic ideas of factor analysis
...
A survey of 169 customers was conducted to assess images
...
These items are shown in Table 14
...
The resulting data set would then be a
matrix of 169 rows (respondents) by 14 columns (semantic differential scales)
...

Figure 14
...
1

Bipolar Dimensions Used in Semantic Differential Scales for Grocery Chain Study
Inconvenient location—Convenient location
Low-quality products—High-quality products
Modern—Old-fashioned
Unfriendly clerks—Friendly clerks
Sophisticated customers—Unsophisticated customers
Cluttered—Spacious
Fast check-out—Slow check-out
Unorganized layout—Organized layout
Enjoyable shopping experience—Unenjoyable shopping experience
Bad reputation—Good reputation
Good service—Bad service
Unhelpful clerks—Helpful clerks
Good selection of products—Bad selection of products
Dull—Exciting

365
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

IDENTIFYING THE FACTORS
If we now input the raw data into a factor analysis program, correlations between the
variables are computed, as is the analysis
...
2
...
It may be helpful to
characterize each of the 14 original variables as having an equal single unit of variance that is
redistributed to 14 underlying dimensions or factors
...
In factor analysis, the analysis first determines how
many of the 14 underlying dimensions or factors are common, and then the common factors are
interpreted
...
2 Some Concepts and Definitions of R-Type Factor Analysis
Factor Analysis:

A set of techniques for finding the underlying relationships between many
variables and condensing the variables into a smaller number of dimensions
called factors
...


Factor Loading:

The correlation between a measured variable and a factor
...


Factor Score:

A value for each factor that is assigned to each individual person or object from
which data was collected
...

2

Communality (h ):

The common variance of each variable summarized by the factors, or the
amount (percent) of each variable that is explained by the factors
...


Eigenvalue:

The sum of squares of variable loadings of each factor
...
e
...


Table 14
...

Principal Components, one of the alternative methods of factor analysis, is a method of factoring
which results in a linear combination of observed variables possessing such properties as being
orthogonal to each other (i
...
, independent of each other), and the first principal component
represents the largest amount of variance in the data, the second representing the second largest,
and so on
...
For a more detailed discussion of the alternative
methods, see Kim and Mueller (1978a, 1978b)
...

Computed as the sum of the squared correlations between the variables and a factor, the
eigenvalues are a measure of the variance associated with the factor
...
2 are a measure of the redistribution of the 14 units of variance from the 14 original
variables to the 14 factors
...
Smith and Gerald S
...
5 percent) of the variance in the original variables
...
2, a scree plot depicts the
rapid decline in variance accounted for as the number of factors increase
...
It is a useful visual tool for determining the number of significant
factors to retain
...

Figure 14
...
0, then the factor is “common” and a part of the solution
...
When we
observe an eigenvalue less than 1
...

Table 14
...
448
1
...
245
1
...
875
0
...
62
0
...
478
0
...
326
0
...
242
0
...
917
11
...
89
7
...
247
5
...
429
3
...
413
3
...
329
1
...
729
1
...
917
49
...
689
66
...
763
77
...
312
86
...
472
92
...
932
96
...
646
100

Table 14
...
If each factor loading in each column were squared, the sum would equal the eigenvalue
shown in Table 14
...
Squaring the loadings (h2) and summing across the columns results in the
amount of variance in the variables that is to be explained by the factors
...

367
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Interpreting the Factors
The interpretation of the factors is subjectively based on the pattern of correlations
between the variables and the factors
...

In factor analysis two solutions typically are obtained
...
e
...
The second solution is accomplished through
rotation aimed at getting loadings for the variables that are either near one or near zero for each
factor
...
This rotation maximizes the variance of a column of the factor loading
matrix, thus simplifying the factor structure and making the factors more interpretable
...
3, Factor 1 is identified by four variables
...
”
We might interpret this factor as the construct up-to-date quality products
...
” This factor might be interpreted as the fast and exciting for
sophisticated customers
...
” One interpretation of this factor is that it represents the
construct of friendliness of store
...
These all
might be a reflection of satisfaction with the shopping experience
...
799 -0
...
184
0
...
692
Quality of products
0
...
033
0
...
081
0
...
724 -0
...
096
0
...
702
Modern
-0
...
216 -0
...
221
0
...
235
0
...
139
0
...
689
Check-out
0
...
720 -0
...
326
0
...
284
0
...
227
0
...
581
Cluttered
0
...
162
0
...
052
0
...
323 -0
...
742
0
...
681
Friendliness of clerks
0
...
298
0
...
433
0
...
013
0
...
011
0
...
587
Service
-0
...
339 -0
...
588
0
...
281 -0
...
290
0
...
634
Shopping experience
-0
...
448 -0
...
552
0
...
3

The example of Table 14
...
However, the loadings are often in the
...
8 range, questioning at what level the variables
make significant enough input to warrant interpretation in the factor solution
...
Smith and Gerald S
...
If the sample size is small,
correlations should be high (generally
...
But as
the sample size increases, the meaning of correlations of lower value may be considered
(generally
...

Overall, it should be obvious that more than one interpretation may be possible for any
given factor
...

This may or may not be a problem, depending upon the objective of the factor analysis
...

One use of factor analysis is to identify those variables that reflect underlying dimensions or
constructs
...
This ensures that all
underlying or latent dimensions are included in the analysis
...
Simply stated, we would like to know how
each respondent scores on each factor
...

Most factor analysis computer programs produce these factor score summates and merge
them with the original data file
...
In short, factor scores (rather than original data values) can be used in subsequent
analysis
...
Canonical
correlation, as we have just seen, examines the relations between two sets of continuous
variables; correspondence analysis examines the relations between the categories of two discrete
variables
...
This analysis
develops ratio-scaled interpoint distances between the row and column categories that depict
accurate and useful positioning maps
...
In strategic terms, the marketing researcher may want to identify (a)
closely competitive brands, (b) important attributes, (c) how attributes cluster together, (d) a
brand’s competitive strengths, and most importantly (e) ideas for improving a brand’s
competitive position (Whitlark & Smith, 2001)
...
Smith and Gerald S
...
1), the main purpose of correspondence analysis is
twofold:
1
...

2
...
This helps interpretation
...

Figure 14
...
Categories with similar distributions will be placed in
close proximity to each other, while those with dissimilar distributions will be farther apart
...

An example from Whitlark and Smith (2004) will be used to illustrate this technique
...
The respondent may
have little or no knowledge of some brands, and providing many ratings can be an onerous task
that produces poor quality data
...
If the list of brands is large, the respondent will be asked to only evaluate a
subset of the list that contains familiar brands
...
This type of question
produces “pick k of n” associative data, where k represents the number of attributes a respondent
associates with a brand and n represents the total number of descriptive attributes included in the
survey
...
9 describes 12 companies
providing communications, logistics consulting, and software support to a group of international
freight handlers and shippers
...
The 12 companies were evaluated by nearly 800 freight
370
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

handlers, who indicated which attributes (pick k of n) best described the companies
...

The results of this analysis show that Oracle and Nokia are perceived as being the most
innovative and as industry leaders, while FedEx offers a relevant and total solution
...
A more
detailed explanation of this technique will be found in the excellent works by Clausen (1998),
Greenacre (1993), and Carroll, Green, and Schaffer (1986; 1987)
...
However rather than focus on the similarity of variables as in factor analysis, the usual
objective of cluster analysis is to separate objects (or people) into groups such that we maximize
the similarity of objects within each group, while maximizing the differences between groups
...
Cluster
analysis can also be used to (a) investigate useful conceptual schemes derived from grouping
entities; (b) generate a hypothesis through data exploration; and (c) attempt to determine if types
defined through other procedures are present in a data set (Aldenderfer & Blashfield, 1984)
...
As such, it seeks to
describe so-called natural groupings, as described in Exhibit 14
...

EXHIBIT 14
...
To illustrate, consider a financial services company that wanted to do a segmentation study among
its sales force of dealers/agents (Swint, 1994/1995)
...
The desire was to profile the dealers/agents and segment
them with respect to motivations, needs, work styles, beliefs, and behaviors
...
The six clusters were then subject to discriminant analysis to how well
the individual clustering attributes actually discriminated between the segments
...


The type of clustering procedure that we shall discuss assigns each respondent (object) to
one and only one class
...
Thus in cluster analysis, we assume here that the underlying structure of the data
involves an unordered set of discrete classes
...

Primary Questions
Clustering procedures can be viewed as preclassificatory in the sense that the analyst has
not used prior information to partition the objects (rows of the data matrix) into groups
...
Moreover, the partitioning is not performed a priori
but is based on the object similarities themselves
...
Smith and Gerald S
...
This type of presupposition is different from the case in discriminant analysis (discussed in
Chapter 13), where groups of objects are predefined by a variable:
Most cluster-analysis methods are relatively simple procedures that are usually not supported
by an extensive body of statistical reasoning
...

Different clustering methods can (and do) generate different solutions from the same data set
...


Given that no information on group definition in advance, we can identify four important
considerations in selecting (or developing) cluster analysis algorithms
...


What measure of inter-object similarity is to be used, and how is each variable to be
weighted in the construction of such a summary measure?
2
...
After the classes have been formed, what summary measures of each cluster are appropriate
in a descriptive sense—that is, how are the clusters to be defined?
4
...
The concept of similarity
always raises the question: Similarity with respect to what? Proximity measures are viewed in
relative terms—two objects are similar, relative to the group, if their profiles across variables are
close or if they share many aspects in common, relative to those which other pairs share in
common
...
The choice of which
objects and variables should be included in the analysis, and how they should be scaled, is
largely a matter of the researcher’s judgment
...
Generally speaking, these measures fall into two classes: (a) distance-type measures
(Euclidean distance); and (b) matching-type measures
...
4
...
4 A Simple Example of Cluster Analysis

372
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

We can illustrate cluster analysis by a simple example
...
Groups will be formed based on two variables, the number of
men who have borrowed money (X1) and the number of women who have borrowed money (X2)
...

We use a proximity measure, based on Euclidean distances between any two branches
...
The first cluster is formed by finding the midpoint between branches 2 and 10
and computing the distance of each branch from this midpoint (this is known as applying the nearest-neighbor
algorithm)
...
The other clusters
are formed in a similar manner
...


Selecting the Clustering Methods
Once the analyst has settled on a pairwise measure of profile similarity, some type of
computational routine must be used to cluster the profiles
...
Each
clustering program tends to maintain a certain individuality, although some common
characteristics can be drawn out
...
Dimensionalizing the association matrix
...
Clusters are then developed on the basis of grouping objects according to their pattern
of component scores
...
Nonhierarchical methods
...
Sequential threshold
...
Then a new cluster center is selected and the
process is repeated for the unclustered points, and so on
...
)
b
...
This method is similar to the preceding method, except that several cluster
centers are selected simultaneously and points within a distance threshold level are assigned to
the nearest center; thresholds can then be adjusted to admit fewer or more points to clusters
...
Optimizing partitioning
...

3
...
These procedures are characterized by the construction of a hierarchy or
tree-like structure
...
At the
next level the two closest points are placed in a cluster
...
Eventually all points are grouped into one larger cluster
...
At the beginning the points are
partitioned into two subsets based on some criterion measure related to average within-cluster
distance
...

373
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

While the above classes of programs are not exhaustive of the field, most of the more
widely used clustering routines can be classified as falling into one (or a combination) of the
above categories
...
The fact remains, however, that even the optimizing approaches
achieve only conditional optima, since an unsettled question in this field is how many clusters to
form in the first place
...
For example,
companies are often interested in determining how their products are positioned in terms of
competitive offerings and consumers’ views about the types of people most likely to own the
product
...
4 shows the result of a hypothetical study conducted
for seven sport cars, six types of stereotyped owners, and 13 attributes often used to describe
cars
...
4

Complete-Linkage Analysis of Product-Positioning Data

The inter-object distance data were based on respondents’ degree-of-belief ratings about
which attributes and owner types described each car
...
The complete linkage algorithm starts by
finding the two points with the minimum Euclidean distance
...
Looking first at the four large clusters, we note the car groupings:
Mazda Miata, Mitsubishi Eclipse
VW Golf
Mercedes 600 SL, Lexus SC, Infinity G35
Porsche Carrera
374
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

In this example, the Porsche Carrera is seen as being in a class by itself, with the
attributes high acceleration and high top speed
...

Studies of this type enable the marketing researcher to observe the interrelationships
among several types of entities—cars, attributes, and owners
...
For example, it can be applied to alternative advertisements, package designs, or
other kinds of communications stimuli
...

Similarly, in the case of consumer packaged goods, the respondent could be shown
alternative package designs and asked for degree-of-belief ratings that the contents of the
package possess various features
...
Hence, four (or more) classes of items could be
represented as points in the cluster analysis
...

Often a useful starting point is to work from a categorization schema of potential foreign
markets
...

To illustrate, we use the study by Green and Larsen (1985)
...
The specific
variables used were (a) growth in Gross Domestic Product; (b) literacy rate; (c) energy
consumption per capita; (d) oil imports; and (e) international debt
...

Table 14
...
T
...
, Export Markets and Economic Change, 1985 (working paper)
...


Clustering was accomplished by use of a K-means clustering routine
...
Smith and Gerald S
...
A total of five clusters was derived based on the distance between
countries and the centers of the clusters across the five predictor variables
...
A
smaller number of clusters led to a substantial increase of within-cluster variability, while an
increase in the number of clusters resulted in group splitting with a minimal reduction in
distance
...
4
...
Most
analysis packages have one or more routines
...
Finally, some academicians have developed their
own cluster routines that they typically make available to other academicians for no charge
...
In this approach
one represents psychological dissimilarity as geometric distance
...

Figure 14
...
S
...
Geographic locations of ten U
...
cities
II
...
S
...
Nonmetric (Rank Order) Distance Data

IV
...
Smith and Gerald S
...
In particular, we show how
MDS takes a set of distance data and tries to find a spatial configuration or pattern of points in
some number of dimensions whose distances best match the input data
...
5
...
S
...
The actual
intercity distances are shown in Panel II
...
For example, we could
use the map to find the distance between Atlanta and Chicago by (a) projecting their points on
axis 1 (East-West), finding the difference, and squaring it; (b) projecting their points on axis 2
(North-South) and doing the same; and then (c) taking the square root of the sum of the two
squared differences
...
However, the converse is not so easy
...

Suppose that we are shown Panel II of Figure 14
...
The task is to work backward and develop a spatial map
...
This is the problem of metric MDS
...
We can build such a data matrix by taking some order-preserving transformation in Panel II
to produce Panel III
...
Then we could apply the same rules and rank order the
remaining 44 distances up to rank 45 for the distance (2,734 miles) between Miami and Seattle
...

In this example where the actual distance data is considered, it turns out that metric MDS
methods can find, for all practical purposes, an exact solution (Panel I)
...

Panel IV shows the results of applying a nonmetric algorithm to the ranks of the 45
numbers in Panel III
...

We should quickly add, however, that neither the metric nor nonmetric MDS procedures
will necessarily line up the configuration of points in a North-South direction; all that the
methods try to preserve are relative distances
...
Smith and Gerald S
...
None of these
operations will change the relative distances of the points
...
In MDS we assume that
individuals act as though they have a type of “mental map”, (not necessarily visualized or
verbalized), so that they view pairs of entities near each other as similar and those far from each
other as dissimilar
...

We assume that the respondent is able to provide either numerical measures of his or her
perceived degree of dissimilarity for all pairs of entities, or, less stringently, ordinal measures of
dissimilarity
...

This model does not explain perception
...
Thus, MDS models are representations of data rather than
theories of perceptual processes
...
Accordingly, it seems useful to provide a
set of descriptors by which the methodology can be classified
...
Mode: A mode is a class of entities, such as respondents, brands, use occasions, or attributes of a
multiattribute object
...
Data array: The number of ways that modes are arranged
...
Hence, in this case, we have one mode, two-way data on judged dissimilarities of pairs
of brands
...
Type of geometric model: Either a distance model or a vector or projection model (the latter
represented by a combination of points and vectors)
...
Number of different sets of plotted points (or vectors): One, two, or more than two
...
Scale type: Nominal-, ordinal-, interval-, or ratio-scaled input data
...
Single mode, two-way data are illustrated by input matrices that are
square and symmetric, in which all distinct pairs of entities (e
...
, brands) in a I I matrix are
often judgment data expressing relative similarity/dissimilarity on some type of rating scale
...
Typically, only I(I – 1)/2 pairs are evaluated
and entered into the lower- or upper-half of the matrix
...
MDS solutions based on single mode, two-way
input data lead to what are often called simple spaces—that is, a map that portrays only the set of
378
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

I points, as was shown in Figure 14
...
Pairs of points close together in this geometric space are
presumed to exhibit high subjective similarity in the eyes of the respondent
...
A set of I judges provide preference ratings of J brands
2
...
The frequency (across respondents) with which J attributes are assumed to be associated with I
brands
4
...
The frequency (across respondents) with which each of J use occasions is perceived to be
appropriate for each of I brands
6
...


These geometric spaces are often called joint spaces in that two different sets of points
(e
...
, brands and attributes) are represented in the MDS map
...
6 we observe that
brands are positioned in 3-dimensional space as points and the attributes are vectors extending a
distance of 1 unit away from the origin
...
The further out on the vector the brand projects,
the stronger the association
...

Figure 14
...
However, in the case of two-mode, two-way data,
the two sets of entities might each be represented as points or, alternatively, one set may be
represented as points while the other set is represented as vector directions
...

The point-point type of two-mode, two-way data representation is often referred to as an
unfolding model (Coombs, 1964)
...
Brand points that are near a respondent’s ideal point are assumed to be highly
379
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

preferred by that respondent
...
Brand to- brand
similarities may be analogously inferred, based on the relative closeness of pairs of brand points
...
5)
...

This ideal point is located at the terminus or end of the vector
...
The farther out (toward vector i’s
terminus) the projection is, the more preferred the brand is for that respondent
...
However, methodologists are developing MDS models that provide more flexibility
than a straight dimensional application
...
Furthermore,
conjoint analysis, to be discussed next, offers high promise for relating changes in the physical
(or otherwise controlled) aspects of products to changes in their psychological imagery and
evaluation
...
However, more recent research has extended the methodology
to similarities judgments
...
In MDS
studies, there are four most commonly used methods of data collection:
Sorting task
...
The number of groups is determined by the subject during the judgment task
...
Stimuli are presented to subjects in all possible pairs of two stimuli
...

Conditional ranking task
...
Each of the stimuli is in turn presented as the anchor
...
Subjects indicate which two stimuli of combinations of three stimuli
form the most similar pair, and which two stimuli form the least similar pair
...
364)
...
They showed that when collecting data,
conditional rankings and triadic combinations should be used only if the stimulus set is relatively
small, and in situations where the maximum amount of information is to be extracted from the
respondents
...
Which of these two to use will depend on characteristics of
the application, such as number of stimuli and whether or not individual-level perceptual maps
are desired
...
Smith and Gerald S
...

Illustrations include (a) choosing a slogan for advertising a soft drink, (b) the relationship
between physical characteristics of computers and perceptions of users and potential users, (c)
effectiveness of a new advertising campaign for a high-nutrition brand of cereal, (d) positioning
in physicians’ minds of medical magazines and journals, and (e) positioning of new products and
product concepts
...

Current research activity in MDS methods, including the increasing use of
correspondence analyses for representing nominal data (Hoffman & Franke, 1986; Carroll,
Green, & Schaffer, 1986; Whitlark & Smith, 2003), shows few signs of slowing down
...
The gap between theory and practice appears to be widening
...


FUNDAMENTALS OF CONJOINT ANALYSIS
Conjoint analysis is one of the most widely used advanced techniques in marketing
research
...
When using this technique the researcher is concerned with
the identification of utilities—values used by people making tradeoffs and choosing among
objects having many attributes and/or characteristics
...
In this chapter, two of the
most popular methodologies are discussed: the full-profile and self-explicated models
...
The stimuli to be presented to the respondent are often designed
beforehand according to some type of factorial structure
...
As an example, a respondent might be
presented with a set of alternative product descriptions (automobiles)
...
When choice alternatives are presented, choice or preference evaluations are made
...
e
...
Once the utilities are determined for all respondents, simulations are run to
determine the relative choice share of a competing set of new or existing products
...
Managers demand models that define products with increasingly more stimulus
attributes and levels within each attribute
...

One early conjoint data collection method presented a series of attribute-by-attribute (two
attributes at a time) tradeoff tables where respondents ranked their preferences of the different
combinations of the attribute levels
...
Smith and Gerald S
...
The
two-factor-at-a-time approach makes few cognitive demands of the respondent and is simple to
follow
...
Moreover, respondents often lose their
place in the table or develop some stylized pattern just to get the job done
...

For the last 30 years, full-profile conjoint analysis has been a popular approach to
measure attribute utilities
...
Each product profile is designed as part of a fractional factorial
experimental design that evenly matches the occurrence of each attribute with all other attributes
...

A third approach, Adaptive Conjoint Analysis, was developed to handle larger problems
that required more descriptive attributes and levels
...
Early in the
interview, the respondent is asked to eliminate attributes and levels that would not be considered
in an acceptable product under any conditions
...
The choice pairs are presented in an order that
increasingly focuses on determining the utility associated with each attribute
...
This choice activity simulates
an actual buying situation, thereby giving the respondents a familiar task that mimics actual
shopping behavior
...
Rather, the conjoint factors
and levels are presented to respondents for elimination if not acceptable in products under any
condition
...
Finally, the relative
importance of attributes is derived by dividing 100 points between the most desirable levels of
each attribute
...
This is done without the
regression analysis or aggregated solution required in many other conjoint approaches
...

Most recently, academic researchers have focused on an approach called Hierarchical
Bayes (HB) to estimate attribute level utilities from choice data
...
This approach again allows more attributes and levels to be
estimated with smaller amounts of data collected from each individual respondent
...
In the
nonmetric version of conjoint analysis, the dependent (criterion) variable represents a ranking of
382
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

the alternative profiles and is only ordinal-scaled
...

The multiple-factor approach illustrated in Figure 14
...
The details of each card are shown on the
left side of Figure 14
...

Figure 14
...
"
3="Professional Gardeners (Horticulture
...
2:1 over Benadryl"
2="Relief 2x longer than Benadryl"
3="Leading long acting Over The Counter "

Legend:
Profile 1 has levels 0,0,0,0:
No Medication is More Effective
Most Recommended by Allergists
Less Sedating than Benadryl
Won’t Quit on You

Column 4: "Gardening",4 Levels
0="Won't quit on you"
1="Enjoy relief while garden"
2="Brand used by millions"
3="Relieves allergy symptoms"

All card descriptions differ in one or more attribute level(s)
...
8) into three piles (with no need to place an equal number in each
pile) described in one of three ways:
Definitely like
Neither definitely like nor dislike
Definitely dislike

The criterion variable is usually some kind of preference or purchase likelihood rating
...
By means of this two-step procedure, the
full set of 16 cards is eventually ranked from most liked to least liked
...

Again, the analytical objective is to find a set of part-worths or utility values for the
separate attribute (factor) levels so that, when these are appropriately added, one can find a total
383
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

utility for each combination or profile
...

While the two-factor-at-a-time and the multiple-factor approaches, as just described, assume
only ranking-type data, one could just as readily ask the respondent to state his or her preferences
on (say), an 11-point equal-interval ratings scale, ranging from like most to like least
...

Figure 14
...
by allergists
Superiority
Less sedating than Benadryl
Gardening
Won't quit on you

Efficacy
No med works faster
Endorsements
Most recom
...
2:1 over Benadryl
Gardening
Brand used by millions

Efficacy
Relief all day
Endorsements
Most recom
...
by allergists
Superiority
Leading long acting OTC
Gardening
Enjoy relief while garden

Card 5

Card 6

Card 7

Card 8

Efficacy
No med more effective
Endorsements
Most recom
...
2:1 over Benadryl
Gardening
Enjoy relief while garden

Efficacy
No med works faster
Endorsements
Most recom
...
by pharmacist
Superiority
Leading long acting OTC
Gardening
Brand used by millions

Efficacy
Right Formula
Endorsements
Most recom
...
Gardening Assoc
...
Gardening Assoc
...
Gardening Assoc
...
Gardening Assoc
...
2:1 over Benadryl
Gardening
Relieves allergy symptoms

Card 13

Card 14

Card 15

Card 16

Efficacy
No med more effective
Endorsements
Prof
...
)
Superiority
Leading long acting OTC
Gardening
Relieves allergy symptoms

Efficacy
No med works faster
Endorsements
Prof
...
)
Superiority
Relief 2x longer than Ben
Gardening
Enjoy relief while garden

Efficacy
Relief all day
Endorsements
Prof
...
)
Superiority
Rec
...
Gardeners (Horticul
...
In practice, if more
than six or seven factors are involved, this approach is often modified to handle specific subsets
of interlinked factors across two or more evaluation tasks
...
6
...
9 shows a table of the resulting utility values for each of the attribute levels
derived for one respondent
...
All one needs to do to estimate the respondent’s utility
score for a given concept profile is to add each separate value (the regression coefficient) for
each component of the described combination
...
Smith and Gerald S
...
) For example, to
obtain the respondent’s estimated evaluation of card 1, one sums the part-worths:
Value for: Efficacy “No med more effective” = 0
...
by allergists” = 3
...
00
Value for: Gardening “Won’t quit on you” = 6
...
00

Figure 14
...
Similarly, we can find the estimated total evaluations for the other 15 options and
compare them with the respondent’s original evaluations
...
The
information in Figure 14
...
Moreover,
all respondents’ separate part-worth functions (as illustrated for the average of all respondents in
Table 14
...
g
...
In short, while the
respondent evaluates complete bundles of attributes, the technique solves for a set of partworths—one for each attribute level—that are imputed from the overall tradeoffs
...
6

Average Utilities of All Respondents

LEVEL
Efficacy
Importance
Endorsements
Importance

Importance %
14
...
24

1
Effective
2
...
83

2
Faster
2
...
18

3
Relief
2
...
Garden
2
...
18
Prof
...
41

385
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Superiority
Importance
Gardening
Importance

17
...
37

Less
2
...
39

Recommend
2
...
74

Relief
2
...
58

Leading
2
...
54

These part-worths can then be combined in various ways to estimate the evaluation that a
respondent would give to any combination of interest
...
It is clear that the full-profile approach requires much sophistication in
developing the profiles and performing the regression analyses to determine the utilities
...

Self-Explicated Conjoint Analysis
The development of fractional factorial designs and required dummy regression for each
respondent places a burden on the researcher and respondent alike, especially when the number
of factors and levels require that a large number of profiles be presented to the respondent
...
The self-explicated
model is based theoretically on the multi-attribute attitude models that combine attribute
importance with attribute desirability to estimate overall preference
...
In this
model, Eo, the evaluation of profile for product or service o, is formed by summing the
importance weighted desirabilities of the attributes and attribute levels that make up the profile
...
Next, the list of attribute
levels is presented and each level is evaluated for desirability (0 –10 scale)
...
Using this information,
the attribute importance scores are used to weight the standardized attribute level scores, thereby
producing self-explicated utility values for each attribute level
...

As with the full-profile model, these scores can be summed and simulations run to obtain
a score for any profile of interest
...
An easy to use online implementation of the
self-explicated model is found at www
...
com
...

Conjoint Reliability and Validity Checks
386
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Irrespective of the method used to carry out a conjoint analysis, it is useful to include the
following ancillary analyses: (a) test-retest reliability; (b) a comparison of actual utilities with
those of random respondents; and (c) an internal validity check on model-based utilities
...
The purpose here is to see if the
judgments are highly enough correlated, on a test-retest basis, to justify the analysis of the
respondent’s data
...
8)
...
Their rank order is to be
predicted by the part-worths developed from the calibration sample of 16 combinations
...

Other Models

So far our discussion has centered on the most widely applied conjoint models—a main
effects model using rankings or ratings
...
Interactions occur when attribute
levels combine to provide a differential effect
...
oatmeal cookies or raisin cookies)
...
Specialized computer programs have been
designed to implement them
...


OTHER ASPECTS OF CONJOINT ANALYSIS
The typical sequence that one goes through to implement a conjoint study involves four steps:
1
...

2
...

3
...
These
product profiles are entered into a consumer choice simulator, along with the earlier computed
individual utility functions
...
Use the respondent’s individual part-worth function to compute the utility for each of the
competing profiles in the choice simulator
...
e
...


Use of Visual Aids in Conjoint analysis
Another problem in the application of conjoint measurement is the pragmatic one of
getting fairly complex concepts across to the respondent
...
8 are not only difficult for the respondent to assimilate, but also introduce unwanted
perceptual differences
...

387
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Wherever possible, visual props can help to transmit complex information more easily
and uniformly than verbal description
...
In the course of
preparing the questionnaire, rather complex experimental factors such as overall size and interior
layout, trunk size and fuel-tank capacity, exterior and interior width, and interior spaciousness
and visibility had to be considered
...
10
...
)
Figure 14
...

Strategic Aspects of Conjoint Analysis
The output of conjoint analysis is frequently employed in additional analyses
...
This fosters two additional types of
analyses: (1) market segmentation and (2) strategic simulation of new factor-level combinations
(most often used to test the viability of new product design configurations)
...

In segmentation studies, the respondents are usually clustered in terms of either their
commonality of utility functions or their commonality of importance weights
...

388
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Strategic simulations are also relatively easy to construct from conjoint analysis data by
simply including each individual respondent’s utility function in a computerized-choice model
...

The simulators can employ a variety of consumer-choice procedures, ranging from
having each consumer simply select the alternative with the highest utility to more elaborate
probability of choice rules, where probability is related to utility differences in the set of
alternatives under evaluation
...
g
...
Conjoint analysis has
already been applied to a wide variety of problems in product design, price elasticity of demand,
transportation service design, and the like
...
6 shows a representative list of applications
...

Table 14
...
Automotive styling
2
...
Car batteries
5
...
Cameras

Other Products
1
...
Hair shampoos
3
...
Clothing: sweatshirts, bras
5
...
Copying machines
2
...
Facsimile transmission
4
...
Personal computer design

Financial Services
1
...
Auto insurance policies
3
...
Credit card features
5
...
Car rental agencies
2
...
Medical laboratories
4
...
Toll collection on toll ways

Transportation
1
...
Transcontinental airlines
3
...
Freight train operations
5
...

Researchers estimate that business firms’ use of conjoint analysis entails several thousand studies
each year
...
com), conjoint methodology is easily accessed by any interested user
...
Conjoint methodology has also been extended to encompass use
occasion and situation dependence in a series of dual-conjoint designs, called componential
segmentation
...
Thus, it is feasible
to extend conjoint beyond the simulation stage (where one finds the best of a limited set of
options) to encompass the identification of the best product (or line) over the full set of
possibilities
...
In sum,
389
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

conjoint methodology, like MDS, appears to be moving into the product-design-optimization
arena, a most useful approach from a pragmatic managerial viewpoint
...
For example, the approach
assumes that the important attributes of a product or service can all be identified and that
consumers behave rationally as though all tradeoffs are being considered
...

In short, MDS and conjoint are still maturing—both as techniques that provide
intellectual stimulation and as practical tools for product positioning, segmentation, and strategic
planning
...

The factor-analytic method stressed in this chapter was principal-components analysis
...

The principal-components model was illustrated on a set of data from a study conducted
by a grocery chain
...
In
addition, an application of clustering was briefly described
...
Psychological distance is given a physical distance
representation
...
A variety of applications were described to give the reader some idea of the
scope of the methodology
...
We first discussed the primary ways
of collecting tradeoff data and then showed how such data are analyzed via multiple regression
with dummy predictor variables
...
We next turned to
some illustrative applications of conjoint analysis, including the design of new products and
services
...

This chapter, together with Chapter 13, covers the major multivariate analysis techniques
and has included brief discussions of lesser-used techniques
...
Other advanced, but related procedures such as
moderated regression, multiple-partial correlation, discriminant analysis with covariate
adjustment, factorial discriminant analysis, to name a few, have been omitted from discussion
...

Our coverage of this vast and complex a set of methods is limited in depth as well as
breadth
...
Smith and Gerald S
...
This
grounding will probably embrace three facets: (a) theoretical understanding of the techniques; (b)
knowledge of the details of appropriate computer algorithms for implementing the techniques;
and (c) a grasp of the characteristics of substantive problems in marketing that are relevant for
each of the methods
...
S
...
K
...
Cluster analysis
...

Ball, G
...
, & Hall, D
...
(1964, July)
...

(Working paper
...

Bijmolt, T
...
A
...
(1995)
...
International Journal of Research in Marketing, 12,
363–371
...
D
...
(1980)
...
In M
...
Rosenzweig & L
...

Porter (Eds
...
Palo Alto, CA: Annual Reviews
...
D
...
(1998)
...
In M
...
Birnbaum (Ed
...
Volume 3: Measurement, judgment and decision-making
Carroll, J
...
, Green, P
...
, & Schaffer, C
...
(1986)
...
Journal of Marketing Research, 23, 271–280
...
E
...
Applied correspondence analysis: An introduction
...

Coombs, C
...
(1964)
...
New York: Wiley
...
T
...
L
...
Export markets and economic change
...
Austin: Department of Marketing Administration, University of Texas at
Austin
...
E
...
J
...
M
...
Multidimensional Scaling: Concepts and
Applications, Boston: Allyn and Bacon
...
J
...
Correspondence analysis in practice
...

Hoffman, D
...
, & Franke, G
...
(1986, August)
...
Journal of Marketing Research, 23,
213–217
...
C
...
Hierarchical clustering schemes
...

Kim, J
...
, & Mueller, C
...
(1978b)
...

Beverly Hills, CA: Sage
...
Smith and Gerald S
...
-O
...
W
...
Factor analysis: Statistical methods and practical issues
...

Kruskal, J
...
, & Wish, M
...
Multidimensional scaling
...

PC-MDS Statistical Software, Scott M
...

Sneath, P
...
A
...
R
...
Numerical taxonomy
...
H
...

Sokal, R
...
, & Sneath, P
...
A
...
Principles of numerical taxonomy
...
H
...

Srinivasan, V
...
Surprising robustness of the self-explicated approach to customer
preference structure measurement
...

Whitlark, D
...
(2001, Summer)
...
Marketing Research, 2, 23–27
...
, & Smith, S
...
(2003)
...

Whitlark, D
...
(2004)
...
(Research Paper
...


392
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Chapter 15Chapter 20
PREPARING THE RESEARCH REPORT
Chapter 2 identified the research report as the culmination of the research process
...

The key attributes of a report are completeness and conciseness
...
The report
should be designed and outlined in detail at the outset
...

This chapter provides information and guidelines that may be invaluable in preparing and
presenting reports, both written and oral
...
Consequently, if the project is not effectively presented,
everything done up to that point will be wasted
...

COMMUNICATION
The importance of the research report derives from its purpose of communicating to
others the objectives, processes and results of the research
...

A poor quality presentation of excellent research leaves the impression of poor quality research
...

First, the communication is has an objective and a purpose
...
Third, it
seeks workable understanding and agreement between the parties
...
Finally, there is implicit recognition of the need
for feedback
...
In short, communication is like playing catch with a baseball: it’s not how many
times the ball is thrown, but how many times it is caught
...

One view of the communication process involved in presenting a research report to
management or an outside client is illustrated in Figure 15
...

Following this process will help you to communicate effectively and avoid a minefield of
obstacles like multiple objectives, political agendas, and special interests
...
Each of these factors must be considered
when determining how to best communicate the results of your research
...

393
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 15
...
Feedback for Continued Success

1
...
Give, solicit, receive feedback
B
...
Goodwill, focus, concurrence
2
...
Power, status, success
4
...
Define the research problem
1
...
Evaluate the problem within the external
environment
3
...
Define the audience
1
...
Presentation audience
C
...
Consider Formality
of the Report
A
...
Detail required
C
...
Develop Your Message
A
...
Develop the presentation
C
...
Present with confidence and knowledge
E
...
Select and Organize Your Information
A
...
Select most effective points
C
...
Enhance message with support materials
(PowerPoint, visual aids, numbers, examples)

SOURCE: From Bienvenu, S
...
R
...
Reprinted by permission of Pearson Education, Inc
...


THE RESEARCH REPORT
Written and oral reports can come in all sizes, shapes, and lengths
...
The length of a report on qualitative research, for example, which can run from as
short as eight pages to more than 100 pages, will be influenced by learning garnered, client
requirements, writing style, and formatting (Greenberg, 1999)
...
There is always need for a report to
management
...
Frequently, the simpler reports written in the marketer’s language are most
understood, read, and, thus, acted upon
...
Such a report emphasizes the methods used and underlying assumptions, and
presents findings in a detailed manner
...

There are two key dimensions to the report-writing process: (a) analysis of the data; and
(b) the actual writing
...
Writing takes the learning and expresses it in a
coherent, logical, and succinct way (Greenberg, 1999)
...
According to Jacoby (1997, pp
...
Smith and Gerald S
...
Thus, someone other than the analyst often does the report writing
...
The most
basic criterion is how well the report communicates with the reader
...
The major concerns are that a research
report be complete, accurate, concise, and clear in what is being said
...
These topics are discussed in more detail in Rehart (1994), Fink (2003), and
Bailey (1990)
...
The writing must be done on the level of the reader’s understanding;
this is fundamental! A report may be incomplete because it is too long or too short
...
In the end, completeness is
always defined by the person(s) who will be reading the report and asked to act upon it
...
But it goes further
...
There may have been flaws in research design, measurement and scaling, sampling,
and analysis that led to inaccurate results being presented
...
The preparation of a
research report calls for attention to detail, including attention to the meaning of every word
used, punctuation, and so forth
...
Thus, as we said above, not all information
obtained needs to be reported
...
Conciseness refers to what is
included and how it is included
...
A concise report
is not the same as a brief report
...
In contrast, a concise report may contain any
amount of detail and be very long
...
But it is efficient; that
is, it conveys all the researcher wants to present in the shortest, most direct way
...
Clarity
derives from clear and logical thinking and organization
...
Smith and Gerald S
...
There is always some logical connection between,
say, a topic or idea and the one that preceded it
...
Time order
...

2
...
The relationship among places and locations determines organization
...
Cause and effect
...
A report may stress one or
the other
...
Increasing difficulty
...

5
...
When a basic framework is understood, content can be organized in
recognized categories
...

6
...
Readers grasp differences or similarities more easily when this
structural organization is used
...
Pro and con order
...


A most important rule is that the report be well organized
...
This means that researchers who prepare reports should develop a writing style
conducive to clarity in presentation
...

Most say the same things, but perhaps in different ways
...
1
...
1 Some Rules for Developing a Writing Style
1
...
Such words are clear and specific
...

2
...
Shorter sentences usually are more readable, all other things being
equal
...

3
...
Use declarative (assertion), interrogative (question),
imperative (command), and exclamatory sentences
...
Also, vary structure among simple, compound, complex, and compoundcomplex sentences
...

4
...
Try to build each paragraph around one idea or topic
...
Use active verbs
...
This does not
mean that the passive voice should never be used
...

6
...
Say what is to be said in as few words as possible
...
Use varying means of emphasis
...
g
...

8
...
Often it is difficult to identify what is natural
...

9
...
Do not overestimate the reader’s
understanding and confuse him or her with highly specialized technical jargon
...

10
...
Avoid trying to say too much in too few words
...

396
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

11
...
The way in which words are put together says a great deal about
the writer
...
While it is rare for a person to have complete control of
the tone of his or her writing, it is important to acknowledge that it has many shades: positive or
negative, helpful or indifferent, courteous or impertinent, humble or arrogant, and so forth
...
101–106
...

Four perspectives of report preparation are presented in Exhibit 15
...

EXHIBIT 15
...
All agree that poorly communicated research is useless
...

Perspective #1: Strengthen Your Communications
In order to strengthen written communication, Teresa Wrobel (1990), Vice-President of
Research 100 in Princeton, New Jersey, suggests six things:
1
...


3
...


5
...


7
...
Presume that a reader feels uncomfortable manipulating or
interpreting numbers
...
Numbers should not interrupt the flow of the text, and the reader should not be forced
to perform mental calculations in mid-sentence
...

You don’t have to analyze everything
...
Even statistically significant results do not always merit discussion
...

Write in plain English (or whatever language is being used)
...
Jargon tends to
conceal the main ideas being conveyed and may be incomprehensible to the intended readers, who
are often not researchers
...

Convey the main ideas quickly
...
The front section of a research document should reduce the findings to
their skeletal elements, and include the conclusions and recommendations supported by key findings
...

Keep the report interesting
...
Keep
searching for new ways of expressing points that convey ideas most effectively
...

Some format limitations must be followed
...
However, such limitations should not stifle creativity
...


Perspective #2 Make It Readable

397
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

A slightly different view is that of Howard Gordon (1998), a principal of George R
...
Mr
...


2
...


4
...


6
...


Present tense works better
...
Results and observations expressed in “now“ terms
sound better and are easier to read
...
” Say:
“People like the taste of the juice
...
Clients want to know what people think now
...
Use active voice where possible, which is in most cases
...
” rather than: “It is believed that
...
Use first person
plural, not third person singular
...
And active is easy to read
...

Findings—don’t use the word findings
...
Marketers should use results, conclusions,
or observations—not findings
...
What is it you want the reader to
take away from your report? If somebody could spend only five minutes with your report, what
should that person know?
Use informative headlines, not label headlines
...
” Your reader will not generally welcome this headline:
“Analysis of packaging
...
”
Let your tables and charts work
...
Don’t take a paragraph or a
page to describe what a table already says
...
The purpose of a table or a chart is to simplify
...
Or use your words to offer interpretive comments
...

Use the double-sided presentation whenever possible
...
It simply presents the table on the left side of the open report
...
The double-sided presentation is one of the most
readable report formats
...
The double-sided system
also allows you to use white space effectively
...

Make liberal use of verbatims
...
Use verbatims if you have them
...
James Nelems (1997), President of The Marketing
Workshop in Norcross, Georgia, cautioned us more than 20 years ago on what should be
avoided
...


2
...


The longer the report, the better
...
Any research report—
regardless of sample size, multivariate techniques, or number of questions— should be reducible to a
one-page executive overview
...
Management has both the time and desire to read such reports—and the confidence to accept
...
Some, particularly the multivariate ones, may confuse rather than
enlighten
...
Research is more than asking questions
...
Or if someone doesn’t use a product, just ask him why
he doesn’t use it
...
But wrong answers to wrong questions hardly
constitute analysis
...
Smith and Gerald S
...


5
...


7
...
Too much research text simply repeats
numbers in the accompanying table
...
” This refers to any kind of research, the outcome of which is
a “single number” that purports to provide an easy decision, such as an on-air recall score, a positive
buying intention, or the overall preference in a product test
...
Low brand awareness should be improved, of
course, but what is unaided brand awareness per $1,000 of advertising spent vs
...
Yet the brand objectives—
gain new users or hold current ones—can produce opposite results with the exact same preference
data
...
This includes summary statistics to one decimal place
...
Yet we’ve actually seen in print a comment on “20% of the
sample of five respondents
...
39–41),
a principal of Lynn Greenberg Associates, a Scarsdale, New York, research consulting company
...


Capture your initial thoughts in a research journal as you work
...

2
...

3
...
Once the basic sections have been
written, it is easier to deal with the more challenging areas
...
Be succinct
...

5
...
The first statement of a section should
be an interpretation of what was learned, followed by more details
...
Turns negatives into constructive learning and be clear about their meaning
...
Integrate information from other sources where appropriate to provide more relevant meaning
...
Feature action-oriented conclusions and implications that address all of the study’s objectives, plus
any important ancillary issues
...
Use a visually appealing format that allows the readers to skim relevant issues
...

10
...

11
...


REPORT FORMAT AND ORGANIZATION
Every report prepared for a research project is unique and custom-made
...

Many techniques people use to organize research reports are based on placement of
information to gain emphasis
...
Thus, something to be emphasized should be presented at the
beginning of the message
...
The
executive (or management) summary is placed at the front of the report and the
conclusions/recommendations at the end of the report
...
Smith and Gerald S
...
Exhibit 15
...
Research reports, however, vary in formality; not all components will always be included
...
2 illustrates how the components are affected by a decreased need for formality
...
There are many
instances where another version of format IV may be used, in which an executive summary will
be substituted for the table of contents
...
2

Components of a Report and Formality

I

II

Title Fly Page
Title Page
Letter of Transmittal
Letter of Authorization
Table of Contents
Summary
Report Body
Appendix

III
Title Page
Letter of
Transmittal
Table of Contents
Summary
Report Body
Appendix

Title Page
Table of Contents
Summary
Report Body
Appendix

IV

V

Title Page
Executive Summary
Report Body
Appendix

Title Page
Report
Body

VI
Report Body

DIMINISHING NEED FOR FORMALITY
SOURCE: Robinson, 1969, p
...


Prefatory Pages

Every research report must have a title page, and may have a title fly page preceding it
(only the title appears on a fly page)
...
The table of contents can be as thin as having only the major sections of the
report listed, or as detailed as including all headings and subheadings
...

Some research reports may include letters of transmittal and authorization
...
Often, such
a letter contains material usually found in a preface or a foreword
...
Other
times, the letter of transmittal is relatively short
...
Except in the most formal of
marketing research projects, a letter of authorization need not be included in the report
...
3
...
This summary reduces the essence of the
study—the why, what, how, conclusions, and recommendations—to one or two pages
...
Exhibit 15
...


400
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

EXHIBIT 15
...
A
number of components can be organized and presented within these sections
...

PREFATORY PAGES
I
...
Introduction
(a) What prompted the
undertaking of the project
(b) Who prompted the project
(c) How the problem was defined

XI
...
Table of Contents
(a) Section subtitles and pages
(b) Illustration titles and pages
(c) Graph titles and pages
(d) Figure titles and pages
III
...
Summary of Report
(a) Problem definition and date
(b) What was researched
(c) When
(d) Where
(e) How and with what techniques
(f) Major findings
(g) Recommendations

VI
...
Research Methods
(a) Research design
(b) Data instruments
(questionnaire, records, etc
...
Methodological Limitations
(a) Weaknesses in research design
(b) External events that may have
influenced findings
(c) Errors in research methods
(d) Alternative causes for findings
IX
...
Conclusions and Recommendations
(a) What research findings show
(b) What actions should be taken
(or not taken)

XII
...
Acknowledgments
(a) Names, titles, and
affiliations of
contributors
(b) Contribution of each
contributor
XIV
...
Smith and Gerald S
...
3

Illustrations of Letters of Authorization and Transmittal
A
...
I am pleased to inform you that
your firm has been chosen as the contractor for this study, pending contract negotiation and discussion of certain
aspects of your proposal
...
m
...
Please call Mr
...

I am looking forward to working with you
...
Ickes, Director
EMPLOYMENT & TRAINING DEPARTMENT

B
...
Steven J
...
Ickes:
This letter accompanies our final report for the Lane County Public Opinion Study
...
We trust that this report will provide meaningful direction to the community as a whole and policy makers
within Lane County
...
Included in the body is the introduction, which deals with why the project
was undertaken and just what the problem was
...
The main body of the report
usually ends with the conclusions and any recommendations to be made
...
Smith and Gerald S
...
These may
include detailed tables that provide more depth than the data presented in the body of the report,
and all forms and measurement instruments used in data collection
...

EXHIBIT 15
...

Typically, it will be a one- or two-page exposition
...
Example B is
drawn from a study of market needs and potential for a local motel
...

A
...
A survey of three
separate samples was conducted to explore the market penetration potential of the product
...
On the basis of data collected from the irrigator
sample, it was concluded that the demand for pipe in 2010 would not be greater than that purchased in
2009
...
There is some indication that
a higher price would be paid for ECP relative to CTP
...

On the basis of data collected from the municipality sample, it was concluded that the demand
for pipe in 2010 would decrease from levels demanded in 2009
...
The respondents would not be willing to pay more for
ECP than for other types of pipe
...

On the basis of data collected from the industry sample, it was concluded that demand for pipe
in 2010 would increase relative to demand in 2009
...
The industrial market is not tightly controlled by PVC
...

Research was conducted on the future price levels of steel and oil to enable a comparison
between plastic pipe (PVC) and steel pipe (ECP)
...
In
general, they are expected to increase
...

In general, it is recommended that ECP not be introduced into any of the markets at this time
...
Motel
The focus of this report is the potential for increase in the occupancy at the Duck Inn Motel
...
These markets
include bus tourism, university departments, and general tourism
...
FOCUS GROUP/MAIL SURVEY
UNIVERSITY DEPARTMENTS
...
MAIL SURVEY

On the basis of data collected for the bus tourism market group, the most important thing to
this segment is the attitude of the motel staff towards those traveling in a large group
...

And finally, for the general tourism target market, the target market is generally happy with the
Duck Inn on most of the important choice criteria, but improvements can be made in advertising,
403
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

appearance, and cohesion with the on-site restaurant to increase occupancy rates further
...
Janitorial Supplies
Objectives

This study attempts to achieve several objectives
...
The need for a distribution center in Eugene
2
...
Current market share of leading competitors
2
...
Degree of loyalty to current suppliers
4
...
Common complaints include poor credibility, insufficient
product knowledge, and high-pressure sales tactics
...
Quite to the contrary, genuine compliments for P&R’s business tactics flowed freely from
respondents associated with P&R
...
For this reason alone, it is critically important that Paulsen & Roles actively participate in this
market
...

In order to further penetrate this market, it is necessary to increase local representation
...

Recommendations
Survey respondents offered recommendations to Paulsen & Roles Laboratories:
Aggressively pursue service as your primary competitive advantage
...

Continue with competitive pricing
...
e
...

Establish a distribution center in Eugene with a showroom, thus enabling a more complete
utilization of P&R’s competitive advantage of complete service to the consumer
...

Develop a seminar program for end-users in the Eugene/Springfield metropolitan area
...

Establish a direct-mail campaign aimed primarily toward the janitorial contractor segment
...

Attempt to acquire the Johnson Wax franchise from Scot Supply
...
Smith and Gerald S
...
They are relevant for both data
analysis and the communication of data of results
...
However, when graphical material is used improperly or is of poor
quality, it can be distracting, misleading, incomplete, confusing and erroneous (Wainer, 1984)
...
g
...
Our brief discussion will
concentrate mainly on tables and charts
...
com can prepare report-ready tables and charts directly from the data file
...

Tables
There are many different ways to show numerical information in tabular form
...
Using tables allows the researcher to minimize the amount of
narrative material in the report
...
More detailed and comprehensive tables should be placed in an
appendix
...

2
...

4
...
This should indicate by itself the contents of the table
...
These are the headings for the columns and rows
...
Explanations for particular items are placed here
...
Source note
...


Figure 15
...

Not only are there different ways to show numerical data in a table, there also are
different ways to present these data
...
1 shows a typical way to present survey response
data
...
Multiple
responses from a respondent were possible
...
If the title cannot be short and descriptive as well, then it may be
necessary to show the question asked
...
1

Presenting Survey Response Data

“What role do you play in the purchase of business forms?”
Role played
Market segments
Hospital
Finance
Manufacturing
Retail
Government
Transportation

Specify
25%
29%
18%
10%
39%
28%

Recommend
44%
32%
27%
0%
48%
28%

Purchase
59%
62%
81%
52%
4%
74%

None
17%
11%
3%
37%
8%
6%

Total sample
size
259
164
205
52
213
109

405
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Figure 15
...

This is quite common for the many specialized cross-tabulation packages, and even in online
packages like Qualtrics where, for example, one dependent variable could allow for simultaneous
cross-tabulations and drill downs of the data for multiple demographic variables
...

Exhibit 15
...
Avoid superfluity
...

Use a pair of scale lines for each variable
...
Put tick marks outside the data region
...

Use a reference line when an important value must be seen across the entire graph, but do not let the line
interfere with the data
...

Avoid putting notes, keys, and markers in the data region
...

Overlapping plotting symbols must be visually distinguishable
...

Visual clarity must be preserved under reduction and reproduction
...
Make legends comprehensive and informative
...

When logarithms of a variable are graphed, the scale label should correspond to the tick mark labels
...

Strive for clarity
...

Subject to the constraints that scales have, choose the scales so that the data fill up as much of the data
region as possible
...
Smith and Gerald S
...

Choose appropriate scales when graphs are compared
...

Use a logarithmic scale when it is important to understand the percent change or multiplicative factors
...

Use a scale break only when necessary
...
Do not
connect numerical value on two sides of a break
...

Graphing data should be an iterative, experimental process
...

Many useful graphs require careful, detailed study
...
100–101
...


Charts and Graphs
Charts and graphs are a popular way to present numerical data
...
The overall advantages
are that relationships and structure can be more easily seen and the narrative can be minimized
...
There are many
different kinds of charts
...
At
times, some combinations may be used in the form of dashboards that provide a real-time
overview of a market as the data is continually updated
...
A simple line chart relates two variables, a multiple line chart shows the relationship
between the independent variable and more than one dependent variable, and a stratum line chart
(or stacked line chart) consists of a set of line charts whose quantities are grouped together
...
5a–d illustrates horizontal and vertical line charts using categorical and ratio scales,
as well as in combination with vertical and horizontal bar charts
...
, 1999)
...
5a-d Example Horizontal Line Charts

(c) Single Horizontal Line Chart: Attributes x Performance Rating

(a) Multiple Horizontal Line Chart: Percentage Change Over Time

407
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

(b) Multiple Horizontal Line Chart: Switching Percent x Percent Savings

(d) Multiple Horizontal Line Chart with Horizontal Bars

Bar Chart

A bar chart is another useful type of graph
...
The stacked vertical bar
chart shown in Figure 15
...
Figure 15
...

Figure 15
...
Smith and Gerald S
...

When using this type of notation, the researcher should make sure that he or she has not added
too much information that confuses the reader
...
6c and 15
...
, 1999)
...
7a-d)
...
When more categories are analyzed, tree diagrams may be used
...

Dot Charts
A simple, yet powerful, type of chart is the dot chart
...
We
have already shown such charts by the scatter plots (diagrams) in Chapter 13
...
7

Example Specialty Charts

(a) Quadrant Chart with Threshold Lines

(b) High-Low Chart: Satisfaction by Channel

(c) Heat Map: Analysis of Ad Interest Areas

(d) Surface Map of Neural Network Analysis

Dot charts are used as output from multidimensional scaling (MDS) analyses (see
Chapter 14)
...
8a shows a MDS result for a computer manufacturer where attributes of
computer manufacturers are plotted as dots
...
8c presents utility functions obtained using conjoint analysis of a 33 fractional factorial design
409
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

examining product attributes (see Chapter 14)
...

Although a literal use of this type of chart would show only dots in the body of graph, this need
not be the case
...
8b
...

Figure 15
...
A beer company was interested in knowing something about preferences for
its beer and a number of competitive brands
...
Prior to applying this procedure,
each respondent was classified in one of three ways:

410
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

1
...
Other brand-loyal—drinking a particular (competitive) brand regularly for at least one year
3
...
9
...
9

Example Thurstone Case V Scale

Other Graphic Aids
Sometimes key results from a study can be presented by use of maps
...
Maps are used to
enhance the meaning of data
...
10a,
which shows the geographic boundaries of selling areas
...
10b shows a map of the eight
western states and reports the percent of respondent pipe sellers who sell PIP in each area
...

Another device that can enhance the presentation is a pictorial chart or pictogram
...
Each symbol
represents a specific and uniform amount or value
...
10c
...
10d
...
For example when doubling the size of a threedimensional symbol, if each dimension is doubled the volume is increased eight times
...
Smith and Gerald S
...
10a-e

Selling Area
Area 1
Area 2
Area 3
Area 4
Area 5
Area 6

Use of Maps and Pictograms in Displaying Research Results

Percentage of All
PIP Respondents
Selling in the Area
78%
71%
57%
14%
57%
43%

Percentage of PIP
Sellers With More
Than 5% Sales in the
Area
43%
71%
57%
0%
57%
29%

(a) Use of Maps in Plastic Irrigation Pipe Study: Selling Areas of PIP
Manufacturers

(b) Use of Maps in Plastic Irrigation Pipe Study: PIP Respondent Selling
Areas as a Percentage of All Respondents Reporting Sales in Area

(c) Pictogram Illustration of Number of Customers by Sales Area

(d) Pictogram Illustration: Symbol Proportional to Value

To

Preference
Without Price

Most Prefer
2nd Most Prefer
3rd Most Prefer
Least Prefer

8
12
47
152

6
16
152
46

124
75
9
12

82
116
12
10

(e) Frequencies of Preference for New Package Designs

Pictures or drawings themselves may be useful to show what something looks like
...
A study was done to test four alternative aerosol packages,
which used a sample of 220 housewives
...
10e shows sketches of the four package
designs and frequencies of choice
...
The sketches greatly aid understanding of the differences between the test items
...
Smith and Gerald S
...
It is not too difficult to get
different impressions from graphs of identical data, each of which is technically correct
...
11 illustrates this point
...
11

Effects of Changing an Axis

By stretching or compressing the horizontal or vertical axis, as the case may be, one can
convey different meanings
...
Such
principles as those discussed in this chapter are focused on obtaining a clear vision, clear
understanding, sound scales, and sound overall strategy for presentation
...

The objective behind this oral report is to identify and emphasize the major findings of the study
and to allow the client or manager to ask questions about matters not clear to him or her
...
When
this is the case, the written report consists of an extended management summary (3–4 pages) and
copies of all overhead projector transparencies, photocopies, slides, or PowerPoint slides
presented in the oral report
...
The presenter must adapt to the audience, and this may require having to
defend the results being presented
...
It
does mean, however, that a presenter should be prepared to deal with any and all questions in a
professional, expert, and competent manner
...
Smith and Gerald S
...

When the researcher wants to use visual aids, it is important to keep in mind the difference
between type of visual aid and medium of presentation
...
In addition, actual actions and behaviors can be shown via videos, DVDs, and
personal computers
...
Exhibit 15
...

EXHIBIT 15
...
She's
also aware that precisely the opposite can occur
...
, public relations consultant
...
" Many embrace the values of PowerPoint as a
potent business tool, but there are others who contend that it's a drag on effective interaction — that it
confuses, distorts and even strangles communication
...
Here are 10
ways to use PowerPoint to help make your business look brilliant, not brainless
...
Hold up your end with compelling material
...
However simple and engaging it can be to
build eye catching slides and graphics, bear in mind that PowerPoint isn't autonomous
...
Build a strong PowerPoint program,
but make sure that your spoken remarks are no less compelling
...
"Remember that you are
creating slides to support a spoken presentation
...
Keep it simple
...
Some authorities suggest no more than five words per line and three
to five lines per individual slide
...
"Do
you really need to have everything up on the screen?"
3
...

PowerPoint's lure is the capacity to convey ideas and support a speaker's remarks in a concise manner
...
For the most part, most effective PowerPoint
displays don't overwhelm viewers with too many figures and numbers
...
If you want to emphasize a statistic,
consider using a graphic or image to convey the point
...

4
...

One of the most prevalent and damaging habits of PowerPoint users is to simply read the visual
presentation to the audience
...
PowerPoint works
best with spoken remarks that augment and discuss, rather than mimic, what's on the screen
...

5
...

A well-orchestrated PowerPoint program brings up a new slide and gives the audience a chance to read and
digest it, then follows up with remarks that broaden and amplify what's on the screen
...
"Never talk on top of your slides
...


414
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

6
...

PowerPoint is most effective as a visual accompaniment to the spoken word
...
Not only can that give your audience a visual break, it's also effective to
focus attention on more verbally-focused give and take, such as a group discussion or question and answer
session
...
Use vibrant colors
...
The standard theme colors are developed by graphic artists to be pleasing and not
overbearing
...
Import other images and graphics
...
Use outside images and graphics for variety and
visual appeal, including video
...

9
...

Some people may disagree with me here
...
Unless it is imperative that people follow a handout while you're
presenting, wait until you're done to distribute them
...
Edit ruthlessly before presenting
...
Once you're finished drafting your PowerPoint slides, assume
you're just one of the folks listening to your remarks as you review them
...
Chances are good your overall presentation will be the better for it
...

Source: Jeff Wuorio, Presenting with PowerPoint: 10 dos and don'ts
http://www
...
com/smallbusiness/resources/technology/business-software/presenting-with-powerpoint-10-dosand-donts
...
In order to plan properly for the
report, the researcher must understand some basic concepts of communications
...
Reports need to
be complete, accurate, concise, and clear in presentation
...
The intended audience must be kept in mind at all times
...
Increasingly, marketing research reports seem to be written less
formally, with a major emphasis being placed on the oral presentation
...
Tables,
charts, figures, maps, and other devices were presented
...

Next we turned to the oral report
...
This allows the potential user of the study results to ask
questions and seek clarification of aspects that may be confusing
...
The written report then
consists of an extended management summary and copies of all overhead transparencies used in
the presentation
...
Smith and Gerald S
...
P
...
(1990)
...
New York: Oxford
University Press
...
S
...
The elements of graphing data
...

Faneuil Group, Inc
...
Training Manual
...

Fink, A
...
How to report on surveys
...

Gordon, H
...
J
...
R
...
L
...
) The Dartnell
marketing manager’s handbook
...

Greenberg, L
...
The challenge of qualitative report writing
...

Gutsche, A
...
Visuals make the case
...

Jacoby, W
...
(1997)
...
Thousand Oaks, CA: Sage
...
G
...
Statistical graphics for visualizing multivariate data
...

Nelems, J
...
(1979, January 12)
...
Marketing News,
12, 7
...
J
...
(1994)
...
Los Angeles:
Pyrczak Publishing
...
M
...
Writing reports for management decisions
...
Merrill
...
H
...
Communicating in organizations
...

Semon, T
...
(1998, November 23)
...
Marketing
News, 32, 24, 17
...
R
...
(2001)
...

Upper Saddle, NJ: Prentice Hall
...
T
...
Making successful presentations
...

Wainer, H
...
How to display data badly
...

Witzling, L
...
, & Greenstreet, R
...
(1989)
...
New York: John Wiley & Sons
...
K
...
Research is useless if it’s poorly communicated
...

Wuorio, J
...
http://www
...
com
416
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

APPENDIX A Statistical Tables
Table A
...
Z is the standard normal variable
...
9474 =
...
00

0
...
02

0
...
04

0
...
06

0
...
08

0
...
0
0
...
2
0
...
4
0
...
5000
0
...
5793
0
...
6554
0
...
5040
0
...
5832
0
...
6591
0
...
5080
0
...
5871
0
...
6628
0
...
5120
0
...
5910
0
...
6664
0
...
5160
0
...
5948
0
...
6700
0
...
5199
0
...
5987
0
...
6736
0
...
5239
0
...
6026
0
...
6772
0
...
5279
0
...
6064
0
...
6808
0
...
5319
0
...
6103
0
...
6844
0
...
5359
0
...
6141
0
...
6879
0
...
6
0
...
8
0
...
0

0
...
7580
0
...
8159
0
...
7291
0
...
7910
0
...
8438

0
...
7642
0
...
8212
0
...
7357
0
...
7967
0
...
8485

0
...
7704
0
...
8264
0
...
7422
0
...
8023
0
...
8531

0
...
7764
0
...
8315
0
...
7486
0
...
8078
0
...
8577

0
...
7823
0
...
8360
0
...
7549
0
...
8133
0
...
8621

1
...
2
1
...
4
1
...
8643
0
...
9032
0
...
9332

0
...
8869
0
...
9207
0
...
8686
0
...
9066
0
...
9357

0
...
8907
0
...
9236
0
...
8729
0
...
9099
0
...
9382

0
...
8944
0
...
9265
0
...
8770
0
...
9131
0
...
9406

0
...
8980
0
...
9292
0
...
8810
0
...
9162
0
...
9429

0
...
9015
0
...
9319
0
...
6
1
...
8
1
...
0

0
...
9554
0
...
9713
0
...
9463
0
...
9649
0
...
9778

0
...
9573
0
...
9726
0
...
9484
0
...
9664
0
...
9788

0
...
9591
0
...
9738
0
...
9505
0
...
9678
0
...
9798

0
...
9608
0
...
9750
0
...
9525
0
...
9693
0
...
9808

0
...
9625
0
...
9761
0
...
9545
0
...
9706
0
...
9817

2
...
2
2
...
4
2
...
9821
0
...
9893
0
...
9938

0
...
9864
0
...
9920
0
...
9830
0
...
9898
0
...
9941

0
...
9871
0
...
9925
0
...
9838
0
...
9904
0
...
9945

0
...
9878
0
...
9929
0
...
9846
0
...
9909
0
...
9948

0
...
9884
0
...
9932
0
...
9854
0
...
9913
0
...
9951

0
...
9890
0
...
9936
0
...
6
2
...
8
2
...
0

0
...
9965
0
...
9981
0
...
9955
0
...
9975
0
...
9987

0
...
9967
0
...
9982
0
...
9957
0
...
9977
0
...
9988

0
...
9969
0
...
9984
0
...
9960
0
...
9978
0
...
9989

0
...
9971
0
...
9985
0
...
9962
0
...
9979
0
...
9989

0
...
9973
0
...
9986
0
...
9964
0
...
9981
0
...
9990

3
...
2
3
...
4

0
...
9993
0
...
9997

0
...
9993
0
...
9997

0
...
9994
0
...
9997

0
...
9994
0
...
9997

0
...
9994
0
...
9997

0
...
9994
0
...
9997

0
...
9994
0
...
9997

0
...
9995
0
...
9997

0
...
9995
0
...
9997

0
...
9995
0
...
9998

A-1
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Table A
...
75


...
95


...
99


...
9995

1
...
816

...
741

...
718

...
706

...
700

...
695

...
692

...
690

...
688

...
687

...
686

...
685

...
684

...
683

...
683

...
679

...
674

3
...
886
1
...
533
1
...
440
1
...
397
1
...
372
1
...
356
1
...
345
1
...
337
1
...
330
1
...
325
1
...
321
1
...
318
1
...
315
1
...
313
1
...
310
1
...
296
1
...
282

6
...
920
2
...
132
2
...
943
1
...
860
1
...
812
1
...
782
1
...
761
1
...
746
1
...
734
1
...
725
1
...
717
1
...
711
1
...
706
1
...
701
1
...
697
1
...
671
1
...
645

12
...
303
3
...
776
2
...
447
2
...
306
2
...
228
2
...
179
2
...
145
2
...
120
2
...
101
2
...
086
2
...
074
2
...
064
2
...
056
2
...
048
2
...
042
2
...
000
1
...
960

31
...
965
4
...
747
3
...
143
2
...
896
2
...
764
2
...
681
2
...
624
2
...
583
2
...
552
2
...
528
2
...
508
2
...
492
2
...
479
2
...
467
2
...
457
2
...
390
2
...
326

63
...
925
5
...
604
4
...
707
3
...
355
3
...
169
3
...
055
3
...
977
2
...
921
2
...
878
2
...
845
2
...
819
2
...
797
2
...
779
2
...
763
2
...
750
2
...
660
2
...
576

636
...
598
12
...
610
6
...
959
5
...
041
4
...
587
4
...
318
4
...
140
4
...
015
3
...
922
3
...
850
3
...
792
3
...
745
3
...
707
3
...
674
3
...
646
3
...
460
3
...
291

df = degrees of freedom
SOURCE: From Fisher, R
...
, & Yates, F
...
Reprinted with permission from Pearson Education, Ltd
...
Smith and Gerald S
...
3

Distribution

df

2
...
01

2
...
05

2
...
90

2
...
975

2
...
995

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
18
20
24
30
40
60
120


...
0100

...
207

...
676

...
34
1
...
16
2
...
07
3
...
07
4
...
14
6
...
43
9
...
79
20
...
53
83
...
00016

...
115

...
554

...
24
1
...
09
2
...
05
3
...
11
4
...
23
5
...
01
8
...
86
14
...
16
37
...
92


...
0506

...
484

...
24
1
...
18
2
...
25
3
...
40
5
...
63
6
...
91
8
...
59
12
...
79
24
...
48
91
...
0039

...
352

...
15
1
...
17
2
...
33
3
...
57
5
...
89
6
...
26
7
...
39
10
...
85
18
...
51
43
...
70


...
2107

...
064
1
...
20
2
...
49
4
...
87
5
...
30
7
...
79
8
...
31
10
...
44
15
...
60
29
...
46
100
...
71
4
...
25
7
...
24
10
...
02
13
...
68
15
...
28
18
...
81
21
...
31
23
...
99
28
...
20
40
...
81
74
...
23

3
...
99
7
...
49
11
...
59
14
...
51
16
...
31
19
...
03
22
...
68
25
...
30
28
...
41
36
...
77
55
...
08
146
...
02
7
...
35
11
...
83
14
...
01
17
...
02
20
...
92
23
...
74
26
...
49
28
...
53
34
...
36
46
...
34
83
...
21

6
...
21
11
...
28
15
...
81
18
...
09
21
...
21
24
...
22
27
...
14
30
...
00
34
...
57
42
...
89
63
...
38
158
...
88
10
...
84
14
...
75
18
...
28
21
...
59
25
...
76
28
...
82
31
...
80
34
...
16
40
...
56
53
...
77
91
...
64

For large degrees of freedom
...
1
...
, by W
...
Dixon and F
...
Massey, Jr
...


A-3
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Percentiles of the F Distribution

Table A
...
10

n

2

1

n

df2/df1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

1

2

3

4

5

6

1

7

8

16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
40
60
120

A-4

9

10

12

15

20

24

30

40

60

120

Table A
...
05
n

n

2

df2/df1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

1

2

3

4

5

1
6

7

8

16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
40
60
120

A-

9

10

12

15

20

24

30

40

60

120

Table A
...
01

n

2= degrees of freedom for denominator

1

n

df2/df1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

1

2

3

4

5

6

1
7

8

16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
40
60
120

A-

9

10

12

15

20

24

30

40

60

120

Table A
...
Smith and Gerald S
...
6
Short Table of Random Normal Deviates
0, = 1 )
–0
...
912
–0
...
257
2
...
970
–1
...
369
1
...
284
–0
...
074
0
...
162
–1
...
323
2
...
761
–0
...
410
2
...
832
–1
...
134
0
...
174
–1
...
298
–0
...
287
0
...
481
0
...
584
–0
...
128
0
...
115
–0
...
330
–0
...
718
–0
...
260
–1
...
619
–0
...
245
0
...
518
1
...
790
0
...
376
0
...
129
0
...
219
2
...
789
0
...
046
–0
...
130
–0
...
884
1
...
210
0
...
719
0
...
558
1
...
600
–0
...
556
–0
...
107
0
...
028
1
...
725
–0
...
379
–1
...
153
0
...
559
0
...
084
1
...
620
–1
...
725
0
...
096
–0
...
855

0
...
343
0
...
003
–1
...
672
–0
...
432
1
...
792
0
...
766
0
...
190
–0
...
284
0
...
562
–0
...
705
0
...
350
–0
...
313
–0
...
575
2
...
143
–1
...
095
–1
...
441
1
...
460
–0
...
921
0
...
738
0
...
642
1
...
548
1
...
182
1
...
226
0
...
493
0
...
523
–0
...
425
0
...
236
–0
...
165
0
...
336
1
...
832
–1
...
901
0
...
628
–0
...
598
0
...
099
1
...
007
1
...
285
1
...
127
–0
...
218
–1
...
278
1
...
811
0
...
384
0
...
570
–0
...
691
0
...
315
0
...
049
–0
...
071
–0
...
473
–0
...
361
–1
...
040

0
...
726
1
...
060
0
...
667
0
...
884
1
...
249
–0
...
361
1
...
057
–0
...
768
–0
...
407
0
...
057
–0
...
260
–0
...
216
–0
...
694
1
...
516
0
...
877
0
...
720
–0
...
426
–0
...
566
0
...
484
1
...
608
–0
...
534
0
...
705
0
...
987
0
...
465
2
...
243
–0
...
161
1
...
337
1
...
696
–1
...
528
1
...
496
–0
...
545
0
...
050
1
...
260
0
...
112
–0
...
698
–1
...
077
2
...
519
–2
...
746
0
...
442
1
...
662
1
...
079
–0
...
725
1
...
457
–0
...
475
0
...
197
–0
...
072
–1
...
489
–1
...
156
0
...
423

0
...
609
–0
...
236
–0
...
481
1
...
539
0
...
621
–0
...
927
0
...
371
–0
...
879
0
...
724
–0
...
043
–1
...
350
0
...
696
0
...
529
1
...
629
1
...
703
0
...
026
–1
...
430
0
...
900
–1
...
950
1
...
720
–0
...
395
–1
...
492
0
...
412
1
...
582
–1
...
607
2
...
716
1
...
090
0
...
351
0
...
278
0
...
033
1
...
151
1
...
629
–0
...
914
–1
...
319
–0
...
688
–1
...
241
–0
...
499
1
...
467
0
...
336
–1
...
476
–0
...
859
1
...
591
–0
...
322
–0
...
286
0
...
420
1
...
325
1
...
267
0
...
480
–0
...
158

–1
...
450
–0
...
442
0
...
842
0
...
497
0
...
472
–0
...
605
2
...
080
0
...
013
0
...
247
–0
...
001
–0
...
289
–0
...
393
1
...
339
–0
...
087
–0
...
725
–1
...
427
–0
...
099
–1
...
264
0
...
192
–0
...
210
–0
...
531
–0
...
673
1
...
052
–0
...
225
0
...
012
–1
...
151
–0
...
676
0
...
017
–0
...
143
–1
...
378
1
...
080
–1
...
221
–0
...
155
1
...
345
0
...
100
1
...
529
0
...
082
–0
...
202
0
...
057
–1
...
945
–0
...
007
0
...
357
0
...
007
0
...
880
–0
...
102
–0
...
119
–0
...
865
0
...
302
0
...
232

A-8
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

GLOSSARY OF TERMS
Accuracy of information Degree to which information reflects reality
...
C
...
C
...

Adequate sample A sample that is of sufficient size to provide confidence in the stability of its
characteristics
...

All-variable regression Multiple regression with all independent variables included
simultaneously
...

Alternate forms reliability The extent to which measurement scores from equivalent forms of
the measure administered to the same sample are the same
...

Ambiguity of questions and responses Errors made in the transmission of information in
interpreting written or spoken words or behavior
Analysis The categorizing, the aggregating into constituent parts, and the manipulation of data
to obtain answers to the research question(s) underlying the research project
...

Applied research See decisional research
...

Arithmetic mean See mean
...

Associative variation A measure of the extent to which occurrences of two variables, or
changes of two variables, are associated
...
”
Attitude measurement Measurement of the person’s mental state toward a certain stimuli
according to specified sets of dimensions and instructions
...

Availability of information Availability of information when a decision is being made
...
g
...

Balancing See matching
...
Smith and Gerald S
...

Basic research See fundamental research
...

Behaviorally-anchored rating scale (BARS) A rating scale using behavioral incidents to define
each position on the rating scale rather than verbal, graphic, or numeric labels
...

Bias The difference between the true value of that which is being measured and the average
value derived from a number of independent measures of it
...

Bounded recall An approach for reducing telescoping by asking questions about events of
concern in previous time periods as well as the time period of research interest
...

Canonical correlation analysis A generalization of multiple correlation analysis to two or more
criterion variables
...

CATI A Computer Aided Telephone Interview in which the interviewer reads the questions from
a computer and enters the responses directly
...
e
...
”
Census All members of the population are included for study
...

Cheater question A question included in a questionnaire that will disclose a respondent’s
giving fabricated answers
...

Classification matrix A cross-tabulation-type table showing the results of discriminant analysis’
ability to correctly classify observations into categories of a criterion variable

G-2
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Cluster analysis A class of statistical techniques whose objective is to separate objects into
groups such that the similarity of objects within each group is maximized while maximizing the
difference between groups
...

Codebook/coding manual A manual which shows how the data have been coded for analysis
...

Coefficient alpha A measure of internal consistency reliability for a multi-item measure which
is a type of mean reliability coefficient for all possible ways to split the items into two groups
...

Coefficient of multiple correlation Correlation coefficient when the number of independent
variables is two or more
...
Collected by commercial
marketing research firms or industry associations
...

Comparative rating scale Objects are rated in comparison with other objects
...

Concept An abstraction formed by generalization about particulars
...
Also known as
constitutive definition
...

Concomitant variation See associative variation
...

Confidence interval A range of values with a given probability of covering the true population
value (parameter)
...

Confusion matrix See classification matrix
...


G-3
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Constant sum question A question in which the respondent is asked to allocate a fixed sum of
points to options to show the importance (or some other attribute or characteristic) of each
option
...

Construct A concept that is the conscious invention of researchers to be used for a special
research purpose
...

Content analysis Coding of free responses to open-end questions
...

Continuity correction A correction factor in chi-square analysis that adjusts for the use of a
continuous distribution to estimate probability in a discrete distribution
...

Continuous panel A sample of individuals, households, or firms from whom information is
obtained at successive time periods
...

Convergent validity A type of construct validity concerned with the correspondence in results
between measuring the same construct by two or more independent methods
...

Conversational interviewing In personal or telephone interviewing the interviewer answers
queries from respondents about question meaning and understanding
...

Correlation coefficient A measure of the association between two or more variables
...

Coverage error Occurs when the sample frame or group from which the sample is drawn does
not represent the population as a whole
...

Criterion validity The extent to which the measurement instrument works in predicting the
future (predictive validity) or reflects the present (concurrent validity)
...

Criterion variable See dependent variable
...
See
Program Evaluation and Review Technique (PERT)
...
Smith and Gerald S
...

Cross sectional design A research design where several groups are measured at the same time,
with each group having been exposed to a different level of the treatment variable
...

Cross validation A procedure in regression analysis and discriminant analysis for examining
whether the predictive equations derived hold up beyond the data on which parameters are
based
...

Data collection techniques The means by which data are collected by communication or by
observation
...
Also known as the basic data
matrix
...

Decisional research Applied research which attempts to use existing knowledge to aid in the
solution of some given problem (s)
...

Depth interview An often unstructured interview that is used to explore the underlying
predisposition, needs, desires, feelings, and emotions of consumers toward products and
services
...

Descriptive study Provides information such as describing market characteristics or functions
on groups and phenomena that already exist; there is prior formulation of specific research
questions
...
”
Determinants of a research project The problem, the researcher, the respondent/subject, and
the client
...
Any event that is necessary and sufficient for the subsequent occurrence
of another event
...

Differential scale A rating scale, assumed to be interval, in which a respondent is asked to
agree with only a subset of items (statements) about an object, with each item having a
predetermined scale value (position on the scale), and the items agreed with correspond to the
respondent’s position on the dimension being measured
...

Direct-judgment rating method A respondent is asked to give a numerical rating to each
stimulus with respect to some designated attribute
...
Smith and Gerald S
...

Discriminant analysis A multivariate technique of analysis that examines the ability of
predictor variables to discriminate (i
...
, to separate) between categories of a criterion
(dependent) variable, or groups
...

Disk-by-Mail A type of survey where the questionnaire is placed on a personal computer
diskette and mail is used to send it to a respondent and by the respondent to return it to the
research organization
...

Distance A characteristic of the real number system where the differences between ordered
numbers are ordered
...

Drop-off/pick-up survey A survey where the questionnaire is left with a respondent and a
representative of the research organization returns later to pick up the completed
questionnaire, or the questionnaire can be returned by mail
...

Electronic surveying See online research
...

Emic vs
...

E-panel Doing panel research using the Internet for data collection
...

Ethical behavior Behavior conforming to professional standards of conduct; it is what most
people in a given society view as being moral, good, or right
...

Ethnography A qualitative approach to research that studies human behavior within a cultural
context
...

Experimental error Noncorrespondence of the true (or actual) impact of, and the impact
attributed to, the independent variable (s)
...
Smith and Gerald S
...

Explicit model A model described verbally, graphically or diagrammatically, mathematically
(symbolically), or as a logical sequence of questions (logical flow)
...

Exploratory study A study whose purposes include the identification of problems, more precise
formulation of problems (including identification of relevant variables), and the formulation of
new alternative courses of action
...

External validity The generalizability of a relationship beyond the circumstances under which it
is observed
...

Factor A variable or construct that is not directly observable but is developed as a linear
combination of observed variables
...

Factorial design A statistical experimental design where there is an equal number of
observations made of all combinations involving at least two levels of at least two variables
...

False positive error Statements by respondents that appear to be complimentary, but really
are not, or when respondents appear to have an attitude and they do not
...

Fixed-size sampling The number of elements to be included in the sample is decided upon in
advance
...
Usually consists of 8 to 12 people
selected purposively
...

Four-group, six-study design Combines an after-only with control group design and a beforeafter with control group design
...


G-7
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Frame error Noncorrespondence of the sought sample to the required sample
...

Framing effects The difference in response to objectively equivalent information depending
upon the manner in which the information is labeled or framed
...
Also known as open-ended text
...

Full profile conjoint analysis Conjoint analysis where different stimulus (e
...
, a product)
descriptions are developed and presented to the respondent for acceptability or preference
evaluations
...

Funnel approach An approach to questionnaire design that specifies a sequence of questions
where one proceeds from the general to the specific or from the easier questions to answer to
those that are more difficult to answer
...

Graphic positioning scale A semantic differential used for multiple object ratings where all
objects are evaluated on each scale item
...

Guided imagery A modified TAT where participants are asked to appraise a product or brand
by concentrating on creating and experiencing an associated image
...

History Events outside an experimental design that affect the dependent variable
...

Implicit model A model that guides a decision but has not been specified in an explicit or
formal manner
...
May be concurrent or
predictive (e
...
, when reported intentions are not carried out)
...

Indexes of agreement Measures of the strength of association between two variables in a
cross tabulation, including the phi correlation coefficient, the contingency coefficient, the
lambda-asymmetric coefficient, and the lambda-symmetric coefficient
...
Smith and Gerald S
...

Information Recorded experience that is useful for decision making; communicated knowledge
which changes the state of knowledge of the person who receives it
...

Instrument effect Changes in the measuring instrument or process that may affect the
measurement obtained in an experiment
...

Interaction The situation in an experiment where the response to changes in the levels of one
treatment variable is dependent on the level of some other treatment variable(s)
...
Some software may customize new questions based on responses to
previously answered questions
...

Internal consistency reliability Reliability within single testing occasions in which the variables
are grouped
...

Internal validity Assesses whether the observed effect is due solely to the experimental
treatments and not due to some extraneous variable(s)
...

Interval scale A measurement scale that possesses the characteristics of order and distance,
and the zero point of the scale is arbitrary
...

Interviewer A person who asks questions in an interview of a respondent
...

Kolmogorov-Smirnov one-sample test A goodness-of-fit test of the agreement between an
observed distribution of a set of sample values and some specified theoretical distribution
Kolmogorov-Smirnov two-sample test A test of whether two independent samples come from
the same population or from populations with the same distribution
...

Laboratory experiment An experiment conducted in a controlled laboratory or laboratory-type
setting
...
Smith and Gerald S
...

Leniency error occurs when respondents consistently use the extreme positions on a rating
scale with relatively little use of intermediate scale positions
...
See also summated scale
...

Logit A type of multiple regression analysis where the categorical dependent variable is
assumed to follow a logistic distribution
...

Make or buy decision The decision by a research client whether the research is to be done inhouse (make) or by an outside supplier (buy)
...

Management summary See executive summary
...

Marketing Decision Support System (MDSS) A coordinated collection of data, systems, tools,
and techniques with supporting software and hardware by which an organization gathers and
interprets relevant information from the business and the environment and turns it into a basis
for marketing action
...
Subsystems are marketing research,
internal records, marketing intelligence, and information analysis
...

Marketing research The systematic and objective search for, and analysis of, information
relevant to the identification and solution of any problem in the field of marketing
...

Also known as balancing
...

Mean The point on a scale around which the values of a distribution balance; it is the sum of all
the values divided by the number of respondents
...
Also known as “Laddering” and “Means-End Chain
...
Smith and Gerald S
...

Measurement error The difference between the information obtained and the information
wanted by the researcher; it is generated by the measurement process itself
...

Memory error Inaccuracy in response that occurs when a respondent gives the wrong factual
information because of not remembering an event asked about
...
on the basis of the attribute or characteristic being studied
...
Methods include objectivist, subjectivist, Bayesian, and
phenomenologist
...

Metric multidimensional scaling Multidimensional scaling in which the input data are ratioscaled
...
that measures direct emotional response to most any communication medium
...

MIS activities Discovery, collection, interpretation, analysis, and intra-company dissemination
of information
...

Mode The typical or most frequently occurring value in a distribution
...

Moderator A person conducting a focus group whose job is to direct the group’s discussion to
the topics of interest
...

Multicollinearity A condition in multiple regression analysis where the predictor variables
show very high correlation among themselves
...

Multi-item scale A scale consisting of a number of closely related individual rating scales whose
responses are combined into a single index or composite score or value
...


G-11
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Multiple choice question A question that has at least two fixed alternative response categories
and the respondents can select k out of n choices
...

Multiple regression analysis Regression analysis with two or more independent variables
...

Multistage sampling A multilevel probability sample in which a sample is selected of larger
areas (or groups), and then a sample is selected from each of the areas (groups) selected at the
first level, and so on
...

Multivariate analysis Statistical procedures that simultaneously analyze measurements of
multiple variables on each individual or object under study
...
The
variable of interest has occurred in a natural setting, and the investigator looks at what has
happened
...

Nomogram A graphic instrument for specifying sample size relating allowable error, confidence
level, mean or proportion, and standard deviation
...

Nonmetric multidimensional scaling Multidimensional scaling in which input data are rank
order data (ordinally-scaled), but which output is interval-scaled
...

Nonprobability sample A sample selected based on the judgment of the investigator,
convenience, or by some other means not involving the use of probabilities
...

Nonsampling error All errors other than sampling error that are associated with a research
project; typically is a systematic error but can have a random component
...

Numerical comparative scale A semantic differential used for multiple object ratings where all
objects are evaluated on each scale item using a verbally-anchored numerical scale
...


G-12
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Observation technique Information on respondents’ behavior is obtained by observing it
rather than by asking about it
...

One-on-one interview See depth interview
...

Online research Using the Internet as a mode of data collection
...

Operational definition Assigns meaning to a variable by specifying what is to be measured and
how it is to be measured
...

Ordered-category sorting A respondent assigns (sorts) a set of stimuli into different categories,
which are ordered on the basis of some property
...

Origin A characteristic of the real number series where there is a unique origin indicated by the
number zero
...

Pantry audit A data collection technique whereby a field worker takes an inventory of brands,
quantities, and package sizes that a consumer has on hand
...

Partially structured indirect interview An interview using a predevised set of words,
statements, cartoons, pictures, or other representation to which a person is asked to respond,
and the interviewer is allowed considerable freedom in questioning the respondent to ensure a
full response
...

Pictogram A pictorial chart that depicts data with the help of symbols such as stars, stacks of
coins, trees, facial expressions, caricatures of people, and so forth
...

Planned information Exists when a manager recognizes a need and he or she makes a request
that information be obtained
...


G-13
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Population The totality of all the units or elements (individuals, households, organizations, etc
...

Population specification error Noncorrespondence of the required population to the
population selected by the researcher
...

Postcoding Coding done after the data are collected
...

Practical significance See substantive significance
...

Precision Refers to sampling error and the size of the confidence limits placed on an estimate
...

Predictive validity See criterion validity
...

Pre-experimental design A research design with total absence of control
...

Probabilistic cause Any event that is necessary, but not sufficient, for the subsequent
occurrence of another event
...

Probit A type of multiple regression analysis where the categorical dependent variable is
assumed to be normally distributed
...

Problem-situation model A conceptual scheme that specifies a measure of the outcome(s) to
be achiever, the relevant variables, and their functional relationship to the outcomes(s)
...
See also critical path method
(CPM)
...
to a non-personal, ambiguous situation that he/she is asked to describe,
expand, or build a structure around it
...

Proposition A statement of the relationship between variables, including the form of the
relationship
...
Smith and Gerald S
...

Purposive sampling See judgment sample
...

Quasi-experimental design A controlled experiment design where there is manipulation of at
least one assumed causal variable but there is not random assignment of subjects to experiment
and control groups
...

Quota sample A nonprobability sample in which population subgroups are classified on the
basis of researcher judgment and the individual elements are selected by interviewer judgment
...

Randomized response technique A technique for obtaining information about sensitive
information
...

Rank correlation The correlation between variables that are measured by ranking
...

Ranking Respondents are asked to order stimuli with respect to some designated property
...

Rating A measurement method where a respondent paces that which is being rated along a
continuum or in one of an ordered set of categories
...

Reactive effects of experimental situation Effects that may arise from subjects’ reacting to the
situation surrounding the conduct of an experiment rather than the treatment variable
...

Regression analysis The mathematical relationship between a dependent variable and one or
more independent variables
...


G-15
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Relevancy of information Pertinence and applicability of information to the decision
...

Repeated measures design A research design where subjects are measured more than once on
a dependent variable
...

Repertory grid A partially structured measurement technique that requires the respondent to
compare objects along dimensions that he or she selects
...

Research design The specification of methods and procedures for acquiring the information
needed to structure or to solve problems
...

Research method Experimental or non-experimental; the major difference between the two
lies in the control of extraneous variables and the manipulation of at least one assumed causal
variable by the investigator in an experiment
...

Research process A series of interrelated steps that define what a research project is all about,
starting with problem formulation and ending with the research report
...

Research question States the purpose of the research, the variables of interest and the
relationships to be examined
...

Respondent A person who participates in a research project by responding and answering
questions verbally, in writing, or by behavior
...

Response error The difference between a reported value and the true value of a variable
...

Sample A subset of the relevant population selected for inclusion in a research study
...


G-16
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Sample frame A means of accounting for the elements in a population, usually a physical listing
of the elements, but may be a procedure which produces a result equivalent to a physical listing,
from which the sampled elements are selected
...
g
...

Sampling error Variable error resulting from the chance specification of population from
elements according to the sampling plan
...

Sampling unit A population element which is actually chosen by the sampling process
...

Scanner data Data on products purchased in retail stores that are obtained by electronic
scanning at checkout of the Universal Product Code (UPC); unit and price information are
recorded
...

Secondary information Information that has been collected by persons or agencies for
purposes other than the solution of the problem at hand, and which is available for the project
at hand
...
It is also
a term used for the effect of the selection procedure for the test (treatment) and control groups
on the results of an experimental study
...

Semantic differential A rating procedure in which the respondent is asked to describe a
concept or object by means of ratings on a set of bipolar adjectives or phrases, with the
resulting measurements assumed to be interval-scaled
...

Sequential sample An approach to selecting a sample size whereby a previously determined
decision rule is used to indicate when sampling is to be stopped during the process of data
collection
...

Simple tabulation A count of the number of responses that occur in each of the data categories
that comprise a variable
...

Simulation A set of techniques for manipulating a model of some real-world process for the
purpose of finding numerical solutions that are useful in the real process that is being modeled
...


G-17
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Skewness A measure of a given data distribution’s asymmetry
...

Socioeconomic characteristics The social and economic characteristics of respondents,
including for example, income, occupation, education level, age, gender, marital status and size
of family
...

Standard deviation A measure of dispersion (variation) around the sample mean, it is the
square root of the variance
...

Standard error of the difference The standard deviation of the sampling distribution of the
difference between statistics such as means and proportions
...

Stapel scale An even-numbered balanced nonverbal rating scale that is used in conjunction
with single adjectives or phrases
...

Static-group comparison A quasi-experimental design in which a group exposed to a treatment
is compared to a group that was not exposed
...

Statistical experimental design After-only designs in which there are at least two treatment
levels
...

Statistical power Ability of a sample to protect against the type II error (beta risk)
...

Stepwise regression Multiple regression analysis in which the independent variable explaining
the most variance is sequentially included one at a time
...

Stratified sampling A probability sample where the population is broken into different strata or
subgroups based on one or more characteristics and then a simple random sample is taken from
each stratum of interest in the population
...


G-18
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010

Stub-and-banner table A table that presents one dependent variable cross-tabulated by
multiple independent variables
...

Sufficiency of information Degree of completeness and/or detail of information to allow a
decision to be made
...
See
also Likert scale and multi-item scale
...

Survey A research method in which the information sought is obtained by asking questions of
respondents
...

Syndicated services Information collected and tabulated on a continuing basis by research
organizations for purposes of sale to firms; data are made available to all who wish to subscribe
...

Systematic error See nonsampling error
...

Tabulation The process of sorting data into previously established categories, making initial
counts of responses and using summarizing measures
...

Telephone interview Interviews that are conducted by telephone
...
It may be forward (report it happening more recently than it did)
or backward (reporting it happening earlier than it did)
...

Test of independence A test of the significance of observed association involving twop or more
variables
...

Thematic Apperception Test (TAT) A test consisting of one or more pictures or cartoons that
depict an ambiguous situation relating to the subject being studied, and research subjects are
asked to make up a story about what is happening, or the subject is asked to assume the role of
a person in the situation and then describe what is happening and who the others in the scene
are
...
Smith and Gerald S
...

Thurstone Case V Scaling Based on the Thurstone’s Law of Comparative Judgment, this
method allows the construction of a unidimensional interval scale using responses from ordinal
measurement methods, such as paired comparisons
...

Total study error Sampling error plus non-sampling error
...

Trend design Data are obtained from statistically matched samples drawn from the same
population over time
...

t – test A test of the difference in mans of two groups of respondents that focuses on sample
means and variances
...

Type II error The probability that one will incorrectly accept a null hypothesis, or any
hypothesis
...

Unobtrusive measures Nonreactive measures of behavior, past and present
...

Unstructured interview An interview in which there is no formal questionnaire and the
questions may not be asked in a prearranged order
...

Validity of measurement The extent to which one measures what he or she believes is being
measured
...

Variation in measurement Differences in individual scores within a set of measurements that
may be due to the characteristic or property being measured (the true difference) and/or the
measurement process itself
...
Smith and Gerald S
...

Verbal rating scale A rating scale using a series of verbal options for rating an object
...

Weighting data Procedures used to adjust the final sample so that the specific respondent
subgroups of the sample are found in identical proportions to those found in the poulation
...

Wilks’ lambda In discriminant analysis it is a multivariate measure of group differences over
discriminating variables
...


G-21
Scott M
...
Albaum, An Introduction to Marketing Research, © 2010



---