Professional Documents
Culture Documents
METHODOLOGY
and
DATA ANALYSIS
2ndEdition
Zainudin Hj Awang
1/2
82research
2/2
Preface
i
Contents
1 INTRODUCTION TO RESEARCH 1
1
What is Research?
What is the Importance of Research? 1
The Needs Research 2
The Role of Research in Business 4
The Determinant for Conducting Research Project 4
The Types of Research 7
The Characteristics of Good Research 12
The Basic Ethics in Research 14
Exercises 14
Introduction
Process 1: Establishing the Problem for Research 18
Process 2: Identifying the Appropriate Research Variables 24
Process 3: Forming Research Objectives, Research Questions, and 25
Research Hypotheses
Process 4: Developing the Theoretical Framework for the Study 25
Process 5: Reviewing Literatures Related to the Problem 28
Process 6: Selecting an Appropriate Research Design 38
Process 7: Collecting Relevant Data 46
Process 8: Analyzing Data using Appropriate Statistical Methods 48
Process 9: Prepare the Final Report to the Management 51
Exercises 50
Chapter 1: Introduction
ii
1.1 The Introductory Paragraph 60
1.2 The Problem Statement 60
1.3 The Objectives of the Study 62
1.4 The Research Questions 62
1.5 The Research Hypotheses 63
1.6 The Significance of the Study 67
iii
The Qualitative Data 107
The Quantitative Data 109
Methods of Data Collection 110
Types and Levels of Measurement 112
The Comparison among Four Types of Scales 117
The Classification of Scaling 119
The Categorical Scale (Voting Scale) 119
The Comparison Scale (Ranking Scale) 124
Exercises 127
iv
9 DATA MANAGEMENT 206
206
Data Management Process
The Process to Code a Question 212
Coding the Structured or Close-Ended Question 212
Coding the Unstructured or Open-Ended Question 215
The Codebook 217
Exercises 219
v
Modeling the Multiple Regression for Latent Constructs 272
Modeling the Mediator Variable in a Model 274
Modeling the Moderator Variable in a Model 275
Analyzing Structural Equation Modeling using AMOS Software 275
Analyzing the Measurement Model 276
Evaluating the Fitness of a Measurement and Structural Model 278
The Steps Involved in Validating the Measurement Model 281
Running the Confirmatory Factor Analysis 281
The Schematic Diagram of a Model 282
The Measurement Model Combining all Constructs 282
Assessing the Fitness of a Measurement Model 286
The Assessment of Normality for a Dataset 291
Analyzing the Structural Model 292
Interpreting AMOS Output for a Structural Model 296
Interpreting the output: The Standardized Regression Weights 296
Interpreting the output: The Unstandardized Regression Weights 298
vi
Solution for Exercises in Chapter 6 351
Solution for Exercises in Chapter 7 353
Solution for Exercises in Chapter 8 354
Solution for Exercises in Chapter 9 357
Solution for Exercises in Chapter 10 359
Bibliography
Index
vii
Chapter 1: Introduction to Research
WHAT IS RESEARCH?
Research activities provide the knowledge and skills needed for the manager in
the fast-paced decision-making environments. Research helps in decision-
making because the analysis of data provides the clear picture to the manager
concerning to the underlying problem, the associated factors, and offers the
possible solution. The researcher interprets the results from the analyzed data
and makes a thorough evaluation on the benefits and consequences for every
possible solution. The researcher also proposes certain recommendations for the
next course of action to be taken by the manager. Thus, the manager is able to
make a better decision with the help of research.
Making the right decision is an ideal practice in any business whenever the
organization encounters a problem. A good organization primarily conducts
research to resolve the critical problems surrounding their business such as
competition, customer satisfaction, product innovation, customer complaints, and
new government policies affecting the industry.
series of steps designed and executed with the goal of getting the best solution
to the underlying problem faced by the organization.
Actually, the entire process in which the managers attempt to solve their
problem is the steps in conducting a research project. The research definition in
the previous section suggests that research involved the process of inquiry,
investigation, examination, and experimentation that need to be conducted
systematically, diligently, critically, objectively and in a logical order. Thus, the
results of research would be the findings, which would help the managers to deal
with the real situation.
Therefore, we can define business research as an organized, systematic
inquiry concerning the problem, and undertaken with the purpose of finding the
best solution to a problem. In addition, the research findings should be able to
clarify all ambiguities surrounding the problem. In research the ambiguities
surrounding the problem are addressed in term of research questions. The
research questions are unanswered questions from the management – and the
research is carried out with the intention to address those unanswered questions.
In other words, research would provide answers to all unanswered questions.
Sometimes the manager has many alternatives to choose from in his decision-
making for the company. In this case, the manager requires more information
concerning each alternative in order to make the right choice. In other words, the
manager needs to carry-out a research. However, the determination of the need
for research centers on the following factors:
For example, Era Timur, a company selling stationery items and providing
internet café plans to open a new branch in Kota Bharu. The manager wishes
to investigate the market potential for the business in the area. The relevant
information that would help the manager in making decision are the
characteristic of the population in the area, the number of potential customers
in the area, the number of schools, government departments as well as private
organizations in the area who are using the internet. And not to forget, the
number of existing competitors already operating in the area is also important.
Since this type of data is usually not readily available, the manager has to
carry out a specific research to collect and analyze the required information.
And not to forget, the manager also needs to know the perception of the
population in the area concerning the proposed business.
Therefore, in this particular case, the company does not have to invest in
research. However, some organizations are willing to spend millions of ringgit
Is sufficient time
Yes Is the data available
Yes
available to do for analysis?
research?
No No
The importance
SHOULD NOT CONDUCT RESEARCH
No Is the decision very
critical for the
organization?
No
In research, a distinction is needed not only about the type of research but also
about the most general category into which the implications of research might
fall. What is the distinction? This distinction is a convenient way to clarify the
research activity to shed light on the purpose and importance of the study. The
most basic distinction of research activities are categorized into applied
research (sometimes called action research) and basic research (sometimes
called pure research or fundamental research).
Applied or action research is carried out for the purpose of solving an existing
problem. The researcher conducts an applied research when he wants to study
about the specific problem. For example, the demand for PROTON is
decreasing. The decrease in demand is reflected in the monthly sales of
PROTON cars nationwide. The management is trying to identify the factors that
contribute to the decreasing demand in order to take an appropriate step before
the problem gets worse.
In this case, the research helps the management to get a clearer picture of
what was happening in the automotive industry, the crucial factors behind it
and the correlation of each factor with the existing economic variables such
as interest rate, rate of inflation, and Gross Domestic Product (GDP). After
obtaining the data, the researcher would perform the data analysis and come
out with certain recommendation based on the output. The management can
use the information from the research report to help them in designing a new
marketing program, determine the pricing policy, and adopt certain
promotional strategy so that the downward trend in demand will change to an
upward trend.
b. The study to identify the main factors considered by the customers when
buying health products in the market
One of the purposes for conducting research is to develop and evaluate the
concepts and theories. The researcher conducts basic research to expand the
body of knowledge in the particular area by determining the theoretical
explanation to the underlying problem. This type of research is not for solving
the problem at hand. Rather, the basic research is carried out to develop the
theoretical foundation concerning certain phenomena which take place and to
expand the limits of knowledge based on the findings.
One carries out basic research to confirm that certain phenomena conform to
a certain theory developed by the previous researchers. If the phenomenon
does not conform to the previous theory, then the researcher would propose
a new theory to explain what was happening. Although basic research does
not provide the solution to a specific problem, it does provide a
comprehensive understanding concerning the problem in general. Hence, the
organizations could apply the theory obtained from basic research to handle
their own problems.
by the frontline staff, the reputation of doctors, and the environment in the
hospital to be the most important criteria that contribute to the patient
satisfaction towards the hospital. Hence, whenever the management wants
to improve their customer satisfaction, they should specifically improve the
quality of those factors.
3. Historical Research
Historical research is the study that relates past events to the current events.
One significant difference between historical research and other types of
research is the type of data collected and the method of data collection.
Historical researcher collects data in a chronological order based on the events
occurring in the past.
In carrying out the historical research, firstly the researcher needs to define
clearly the problem to investigate. The problem of a historical research is
unlimited in scope since it consists of the events of the present day and the
events of the past. The past events have become a historical database that
is a vast collection of documents whereby much of them might be difficult to
verify their authenticity. This step is much like any researcher’s mental effort,
which is usually resulting from a personal interest in a particular area.
Interpret theFigure
results 1.2: The Collect
processdata
in Historical
and Research
Formulate relevant
and write report perform analysis research hypotheses
1. It is based on the work of previous researchers in the area. This does not
necessarily mean that the new researcher duplicates their work, but the new
researcher refers the work of others and continues from where they stopped.
In other words, a new researcher continues from where the earlier researcher
finished their work in order to improve what has already been achieved. If the
new researcher does not continue from where they finish, he might be
repeating the same work all over again or he is re-inventing the same wheel!
Researchers should observe and follow the research ethics at all time during
their work. The ethics in research refer to a code of conduct or expected
societal norm of behavior of researchers while conducting research. Ethical
conduct applies to the researchers themselves, the organization involved in the
research project, the organization sponsoring the research, and the
respondents who participate in the research.
3. Coercion - The respondents are participating in the study at their own will.
Their response during data collection should be free from any influence.
Even, the respondents are free to participate or not participate in the study.
6. Sharing the findings – After the study has been completed, the report should
be presented to the audience. The report should also be made available to be
viewed by the respondents who participated in the study.
Exercises
True-False Questions
STATEMENT TRUE FALSE
A research that contributes towards the enrichment of the
1.
body of knowledge is called an action research.
Applied research obtains data to support theories or related
2.
hypothesis about social phenomena.
Researchers should observe the ethics seriously when
3.
conducting their research work.
A good manager is the one who is sensitive to the
4.
information affecting the business.
a. Pure research
b. Basic research
c. Applied research
d. None of the above
2. When we want to expand the limits of knowledge, without directly involving the
solution to any particular pragmatic problems, we conduct ………… research.
a. Basic
b. Applied
c. A telephone interview
d. A focus group
a. Time constraints
b. Availability of the data
c. Availability of company employees who are survey experts
d. Value of business research information in relation to its costs
a. Informed consent
b. Guaranteed anonymity
c. Confidentiality procedures
d. All of the above.
8. A research involves:
a. Segmentation research
b. Product research
c. Pricing research
d. Forecasting research
Essay Questions
1. For each of the situation below, state whether the research should be
descriptive or causal. State the reasons for your choice.
2. Situation A:
A university lecturer is interested to investigate the factors that contribute to
absenteeism in the workplace. After gathering some information, the lecturer
identifies factors such as the flexibility of working hours, the amount of
monthly incentives given, and the working conditions as primary factors
influencing absenteeism.
Situation B:
An advertising agency wishes to conduct a research to find out the
effectiveness of its sales promotion on consumer demand of its product.
iii. A profile study of people who invest in the quick-rich scheme organized
through the internet.
iv. A study on the factors that influence the monthly expenditure of a family
living in Kuala Lumpur.
8. Miss Sarah, the owner of a small business (a woman’s dress boutique) has
invited a consultant to provide report concerning how her business is different
from similar businesses within a 40km radius with respect to the usage of
information technology for advertisement, the pricing strategy, the quality
customer service, level of customer satisfaction, the store appearance, and
also the product choices.
Situation 1:
A researcher investigates five personality traits to see if they can explain
the purchasing behaviour of automobile buyers.
Situation 2:
A researcher examines how the heat processing and the use of flexible
pouches for storage help retain the freshness of fish.
10. What are the differences between applied research and basic research?
INTRODUCTION
The activities involved in the first step bring the researcher and the
management together because the manager needs the researcher to help his
decision making. Usually before the researcher undertakes a consultancy work,
the manager has to state what he believes to be the main issue surrounding the
problem. At this point, the researcher would make sure that the manager
understands the actual problem. The researcher then helps the manager to look
at the problem objectively, and justify whether it is really a problem or it is only a
symptom to a problem. It is important for the manager and the researcher to
agree on the problem to be researched so that the research processes would
provide information which is relevant to the problem.
The researcher should always remember that the research will provide the
required information that helps the manager in making a decision. That is
why the researcher must discuss thoroughly with the manager before
begin to conduct a research. The discussion will reveal the actual problem
faced by the manager and hence what type of information required by the
manager from the research to facilitate his decision making for the
organization. In this case, the researcher and the manager need to agree
on the problem to be researched and the output required in solving the
problem.
The informal interview with the professionals in the field is another method
that helps researcher to understand the research problem. The
professional is the person who has an expertise in the area related to the
study. For example the economists and the bankers are among the
experts who have broad knowledge concerning the economic outlook of
the country. The professional body in this industry is The Association of
Automobile Manufactures (AAM). The informal interview with them would
reveal the clearer scenario regarding the variables affecting the present
economy and the demand for certain products and service in the country.
The opinion from the experts would also help the researcher to identify the
main variables surrounding the problem. Thus, these variables should be
included in the research.
What problem that triggers you to carry out research? The most important
step in problem identification process is to understand the background of the
problem. The failure to understand the actual problem will make research
project begins without clear objectives and direction.
Normally an exploratory study and informal interview with the expert in the
field might help both the researcher and the management to identify the
problem. The researcher should study the similar problem occurring in other
organizations, trying to figure out the similarities with the problem at hand in
terms of the variables involved. For example, the background of the problem
is the stiff competition in terms of product pricing, product range, service
quality, and product quality from foreign importers
The actual problem may be identified only after gathering the background
information and conducting an exploratory study. For example, in measuring
the customer satisfaction towards a certain supermarket, a manager may
think many possible factors that influence customer satisfaction. Among them
are the products, the services, the location, the price, etc. The most critical
task of the researcher is to isolate and identify the most likely causes based
on the symptoms of the problem. However, certain factor that appears to be a
problem may be only symptoms of a deeper problem.
By looking at the symptom, he has an opinion that the customers might prefer
the exclusive water park as compared to his traditional swimming pool.
However, after an exploratory study using the focus group interview with
customers, he found that the actual problem is the change in demographic
variables, the change in lifestyles, and hence the change in the preference of
customers. Children in this neighborhood have grown up and their preference
has switched to the water park. So, there is no demand for traditional
swimming pool from the generation of younger and outgoing lifestyles.
As another example, the decreasing sales volume is the symptom that can be
seen (daily, weekly, monthly, and annually). Nobody knows exactly the actual
problems that cause poor demand from customers until the research is
carried out. The research findings might discover the design of products is no
longer suitable for current lifestyle, and also the pricing package offered by
competitors is much better. These two factors cause customers to switch.
Remember: the manager cannot act on the symptom of the problem
(decreasing sales), but the manager can act on the factors that cause the
problem of decreasing sales (product design and pricing strategy).
The unit of analysis is the most important aspect in research and it is the
focus of the study. The researcher must decide clearly whether the research
will focus on individuals, households, organizations, certain departments
within an organization, certain geographical areas, or certain groups in a
population. Unit of analysis is the focus where the respondents evaluate its
performance which relates directly to the problem to be researched. In the
case of the swimming club, the unit of analysis is the swimming club itself.
The respondents are selected among the customers to evaluate the
performance of the swimming club in terms of its corporate image perception,
its services quality performance by its employees, etc. The respondents are
the existing customers and also the people who used to experience the
service provided by the swimming club.
The research problem is normally associated with many factors, and those
factors are basically represented by variables. A variable is anything that
changes in value and represents a quantity that can differ in values,
magnitude, or strength. In other words, the variable is generally anything that
may assume different numerical values. Among the variables normally
associated with marketing research are perceived quality of product or
service, product image, product price, competitor’s price, socio-economic
status of respondent such as education, occupation, income, etc.
Research objectives set a list of what the research is going to achieve. They
provide evidence of the researcher’s clear understanding about the purpose
and direction of the proposed study. The research objectives should follow
from the definition of the research problems. Formally stated research
objectives provide the guidelines for determining which other steps must be
undertaken. The assumption is that if the objectives are achieved, the
decision maker will have the information needed to solve the problem.
Overall, research objectives are the elements of what your research is trying
to achieve. Most of the time the research objectives begin with the word “to”
1. The objective of this study is to identify the factors that contribute to the
declining demand for Proton among the Malaysian customers.
2. The objective of this study is to generate the items that constitute
service quality in higher education from the students’ perspective.
3. The objective of this study is to measure the strength of association
between socio-economic status of customers and their perception of
quality concerning the service provided by higher educational
institutions in the country.
4. The objective of this study is to assess the influence of service quality
provided by the university and the corporate image of a university on
students’ loyalty towards their university in doing postgraduate study.
Example 2:
Suppose that you are studying the effects of on-job training for
employees on their productivity and commitment towards the
organization.
For the above study you are required to develop the required research
question and its corresponding hypothesis:
Solution:
First of all, you need to determine the variables of main interest in the
study. In this case, the variable of interest is employees’ productivity and
their commitment. This study is trying to investigate the importance of
training on productivity and commitment. Thus, the dependent variables
are productivity and commitment and the independent variable is on-job
training. Now you can formulate the research question and its
corresponding hypothesis. In this case, the independent variable is on-job
training, while the dependent variables are productivity and commitment.
The unit of analysis in this study is an employee of the organization.
Productivity
Training
Commitment
From the theoretical framework, the hypothesis of the study would be:
Example 3:
The independent variables are the quality of food served and also the
quality of service provided by the food operator.
Food
quality
Eating
at dining
hall
Service
quality
This step will take place after the study identifies the appropriate variables
through extensive literature search. The theoretical framework is a schematic
diagram showing how the study believes the variables are related among
each other. From the schematic diagram, the study would debate the
rationale why certain relationships between variables are believed to exist in
reality. The nature and the direction of relationships among the variables are
also discussed and the discussions are supported with findings from past
Prof Dr Zainudin Awang 32
UniSZA
Chapter 2: The Research Process
Service
Quality
Customer Customer
Satisfaction Loyalty
Corporate
Image
Figure 2.1: The theoretical framework relating the four variables in the study
How the variables in the study relate to each other?
In the above diagram, the study believes that service quality performance
provided by the organization and corporate image of that particular organization
are two independent variables. These two variables are believed to influence the
mediating variable namely customer satisfaction. Finally, the study believes that
the satisfied customers would become loyal customers towards the organization
in the future. In this case customer loyalty is the intended dependent variable.
(Remember: The relationship among variables in a study and any argument made
regarding the relationship must be supported by the literatures in that particular field.
That is why literature review is one of the important chapters in any research)
Prof Dr Zainudin Awang 33
UniSZA
Chapter 2: The Research Process
1. To ensure that all important findings in the previous studies are included as a
basis to support the development of current research.
3. To guide the researcher in selecting the variables for the study and to provide
the rationale for selecting those variables.
5. To avoid the researcher from “reinventing the wheel”. The researcher might
be researching exactly what has been researched if he does not review the
literatures prior to doing his research.
6. To share with the reader the findings of previous studies that relate closely to
the current study. Thus, the replicability of the findings of the current research
is enhanced.
7. To provide the framework for establishing the importance of the current study
as well as a benchmark for comparing the findings of the current study. The
researcher can debate as to how his finding is consistent or not consistent
with the results obtained by previous researchers.
(1) Identify the relevant sources for literature in the field of interest,
(2) Obtain and read all literatures which is relevant with the study,
(3) Write up the literature review based on the information obtained.
After making literature search, the researcher might have a long list of
articles from newspapers, magazines, books, and journals. Then, the
researcher can select certain articles from the group. In the selection
process, the researcher should go through the abstract of the articles. The
researcher should consider whether the articles will make useful
contribution to the literature. From the articles, the researcher should
systematically focus on the knowledge gap that is being researched, the
research design employed in the study and finally how the finding
consistence with the theory etc.
In reviewing the content of research studies, the researchers need to
consider what information to be extracted and to be summarized for the
literature review. A good literature review should include the following
elements:
i. Mention about the main problem addressed in the study.
ii. State the central purpose or the main focus of the study.
iii. State information about the population and sample of the study.
iv. Review the main findings of the study (LR) that relate to the current
study.
v. In some cases, the researcher needs to point out the technical
review depending on the forms of literature review (integrative,
theoretical, or methodological review). This process will facilitate
the writing up of the literature review with minimum disruption and
maximum efficiency.
There are several methods of citing references and using quotations for
literature review. The American Psychological Association (APA) offers
detailed information regarding citations, quotations, and references of the
literature.
The research design serves as a master plan of the methods and procedures
for the researcher to collect and analyze the data required by the research.
Determining the most suitable research design is a function of research
objectives and the specific requirements of the research. In this master plan, the
researcher must consider the type of data required, the design technique
(survey, observation, experiment, and case study), the sampling methodology
and procedures, the schedule and the budget.
Example 2.7:
Example 2.8:
and another 10% use more than one product category. The study also
found that 60% of customers who use cosmetics are loyal to certain
brands, and another 40% do not care much about brand. In addition, 70%
of the customers who are brand loyal choose the brand advertised by their
favorite female artist.
For example, does smoking cause cancer? - refers to the question for
causal study.
Rather, are smoking and cancer related? - refers to the correlational
study.
relationship between the causal factors and the effect predicted on the
market performance variable under investigation. This type of research
design allows decision makers to gain the highest level of understanding
in the research process. In addition, understanding the cause-effect
relationships among market factors allow the manager to make if-then
statements about the variables.
There are two fundamental approaches to gathering raw data. One is to ask
respondents on the variables of interest through the arranged interviews or
structured questionnaires. The other is to observe the respondents and record
the variables of interest using professional observers or high-tech devices.
Self-administered questionnaires, personal interviews, internet survey,
telephone interviews, and focus groups discussions are just some of the tools
researchers can use to collect data.
Table 2.1 summarizes the conditions favoring the use of census over
sample. A research project usually has some budget and time constraints. A
census is appropriate when the budget amount is large and the time completion
is long. However, census is unrealistic if the population is large; for instance, the
population of certain consumer products in Malaysia. A census is also preferred
when high variance in terms of the characteristics of population. For example, a
study conducted to compare the usage of Proton cars versus Chevrolet in
Malaysia.
Pilot testing
A pilot testing is the survey conducted prior to the actual field study. The
purpose is to detect weaknesses in questionnaires and its instrumentation so that
correction can be made. From the pilot testing work, the researcher would know
the actual logistic requirement and budget for the actual research so that no
more interruption occurs when the actual research is carried out. Pilot testing is
very important since the researcher might have overlooked many aspects which
would affect the outcome of the research.
In this step, the researcher begins the process of turning raw data into data
structures that can be used in generating meaningful and useful information
for the manager. The data analysis procedures can vary widely in
sophistication and complexity. For example, it can vary from simple frequency
distributions (which produces percentages) to sample statistical measures
(which produces mode, median, mean, range, standard deviation, and
standard error of the sample) to multivariate data analysis techniques (which
produce Anova, Regression, etc).
1. The number of groups involved in the study, whether it is one group, two
groups or more than two groups.
2. If two or more groups are involved, then the next question is: Are they
independent or dependent to each other?
3. The measurement scale employed in data collection, whether the scale is
nominal, ordinal, interval or ratio.
The researcher must transform data structures into information. This step is
where the required information is created for the manager. It is in this step that
the researcher and the manager obtain the results of statistical testing and
creates a narrative description of the results. Narrative interpretation relates to
the integration of several variables into an understandable statement that the
decision maker can use to address the research question.
This section is very important and should be the main focus in any research.
The research does not have a value if the result is not properly presented and
interpreted in the findings. The report should be prepared and presented in
such a way that the manager can assess whether the objectives set earlier are
accomplished. The report presentation also should be able to address the
research questions clearly, so that no more ambiguity concerning the
underlying problem is left after the research project has been completed. You
can obtain a copy of final report of researches done previously in the library.
The most important element in the report is the highlight of information required
by the management to be used in their decision-making. The data analysis, the
interpretation of output, and the presentation of findings should be made based
on what has been agreed upon with the manager before the research project
begins. That is why the thorough understanding of the research process is
important before the researcher undertake any research project. Once the
researcher understands clearly the research process and the task involved, he
can begin writing a formal research proposal to undertake the research.
EXERCISE
True-False Questions
a. A series of steps in identifying the variables to study, and the method for
conducting the study.
b. A step in finding conclusion to any problems.
c. Creating any databases.
d. None of the above.
a. Primary
b. Secondary
c. Quantitative
d. Qualitative
a. Past results
b. The decision maker
c. Theory
d. Industry experts
a. Null
b. Alternate
c. Non directional
d. Syndicated
a. Panel study
b. Cross-sectional study
c. Longitudinal study
d. Tracking study
a. Introduction
b. Methodology
Prof Dr Zainudin Awang 51
UniSZA
Chapter 2: The Research Process
c. Literature review
d. Research finding
a. Problem definitions
b. Theoretical frameworks
c. Analytical models
d. All of the above
a. An analytical model
b. A hypothesis
c. A research question
d. Relevant characteristics
a. Problem definition.
b. Theoretical frameworks.
c. Analytical models.
d. Research questions.
18. When a researcher cannot find any literature to refer to as a basis for his
research, he should carry out ___________ research.
a. An exploratory
b. A conclusive
c. A descriptive
d. A causal
a. Laboratory experiment.
b. Structured questionnaire.
c. Case study.
d. Pilot survey.
20. Which of the following is NOT the task involved in defining research
problems?
Essay question
2) For each research question below, write down the possible null and research
hypotheses:
10) The president of “Mee Segera Sdn Bhd” suspects that half of his 500 male
and female workers are not very motivated to do their job. He also noticed
that those who are motivated are also the ones who have worked for more
than five years with the company. These workers have received salary
increase and have been rewarded accordingly over the years. Suppose you
are asked to conduct the study for this company,
a. State the problem statement for this study. Determine the relevant
variables and illustrate them using a theoretical framework.
b. State five research questions pertaining to the problem of the study.
c. State three research objectives of the study and describe how each
objective can be accomplished through appropriate statistical analyses.
d. Based on research questions and objectives of the study, design a short
questionnaire consisting of the following:
Research proposal contains the detailed activities of how the researcher will
carry-out the proposed research project from the beginning until it is completed.
The research proposal serves as a direction for the researcher in doing his
research work. Doing research without the research proposal is like sailing the
ship without a captain! Sooner or later the ship would lose its direction and the
journey would not reach its final destination.
The general outline regarding the items that should be included when one is
writing a research proposal is presented in Table 3.1.
58
THE CHAPTERS AND SUB CHAPTERS IN ANY RESEARCH PROPOSAL
CHAPTER I- Introduction
1.1 Introductory paragraphs
1.2 Statement of problems
1.3 Objectives of the study
1.4 Research questions
1.5 Research hypotheses
1.6 Significance of the study
59
THE STEP BY STEP GUIDE IN WRITING A RESEARCH PROPOSAL
Chapter 1 – Introduction
The main intention here is to get your readers interested in your study.
Remember, you are selling your idea and you must catch the attention of your
readers and make them be with you all the way. It sets the stage for you to
proceed and elaborate on your subject. The introduction often contains
dramatic and general statements about the need for the research study. You
can use illustrations or quotes to set the tone. The objectives of the
introductory statements are:
1. To make the audience understand the urgent need for a solution to the
problem or the research to determine the right methods to remedy the
existing problem in order to bring the situation back to normal.
2. To highlight what has gone wrong and the potential threat it will bring to
the survival of an organization, and hence a research is required.
60
1.2 The Problem Statement
The statement of problem is the centre attention of your research. It is just one
main sentence and supported by a few paragraphs of elaboration. In your
sentence, you should take an opportunity to:
6. Highlight the literature gap. The gap can either be the specific research
has not been carried out previously, or certain important variables were
not studied in previous researches, or the population of the study has
changed significantly in term of their socio-economic status which requires
a new research, etc.
In short, you must present something critical that requires your proposed
research to be approved in order to ensure the long-term survival of the
organization and to contribute to the literature. Remember, if you fail to prove that
there is a problem, it simply means you fail to justify the need for your research.
61
Example 2a: The Problem Statement
The objective of the study is what you intend to accomplish in the study. You
need to tell your audience what you plan to achieve in solving the problem. If
you have more than one objective, you can list them specifically after your
introductory sentence.
62
1.4 The Research Questions
Research questions are „ambiguities‟ that the researcher would like to address
in his research. In the research, he intends to determine the answer to his
research questions. Research questions often came into researcher‟s mind
after he sets his research objectives. These questions are normally similar to
the questions from the audience regarding the underlying problem. A good
researcher is able to predict the questions that will come from the audience,
and will include these questions as his research questions when he prepares
his proposal. Hence, there is no more ambiguity arise after the research is
completed since all questions have been addressed by the research. If the
research is completed with many unanswered questions, then that particular
research is a failure.
63
Example 5: (A case study)
Therefore, the researcher would measure two variables in his data collection
process namely, the amount of environmental news exposed to customers,
and the consumers‟ intention to purchase certain cosmetic products.
64
Another hypothesis here is concerning the relationship between the level of
consumers‟ education and their level of awareness towards environmental
issues. In other words, the researcher may hypothesize that consumers with
different levels of education may have different perception on environmental
issue, and hence will act differently concerning product selection. Hence his
hypothesis would become:
Here the researcher will measure two variables in his data collection namely,
the level of consumers‟ education, and the level of consumers‟ awareness
concerning environmental degradation.
Again, here the researcher will measure two variables in his data collection
namely, the level of consumers‟ education, and the level of environmental
awareness. The study will measure the strength of relationship between the
two variables.
A researcher may derive a hypothesis from his own observation, his previous
reading of literature, or from his own instinct and intuition regarding the issue.
A shrewd researcher manages to sense something when faced with certain
research questions. Actually, hypothesis facilitates in answering the research
question.
65
Example 6: The Research questions and their corresponding hypotheses
Research Question 1:
How significant is the influence of service quality on customer satisfaction?
Research hypothesis to address RQ 1
H1: Service quality has a significant influence on customer satisfaction.
Research Question 2:
How significant does the influence of service quality on customer loyalty?
Research hypothesis to address RQ 2
H2: Service quality has a significant influence on customer loyalty
Research Question 3:
How significant does the influence of customer satisfaction on customer loyalty?
Research hypothesis to address RQ 3
H3: Customer satisfaction has a significant influence on customer loyalty
Research Question 4:
Which demographic variable, if any, moderates the relationship between service quality
and customer satisfaction?
Research hypothesis to address RQ 4
H1: The respondents‟ age group moderates the relationship between service quality and
customer satisfaction
H2: The respondents‟ gender group moderates the relationship between service quality
and customer satisfaction
H3: The respondents‟ occupation moderates the relationship between service quality
and customer satisfaction
H4: The respondents‟ education level moderates the relationship between service
quality and customer satisfaction
H5: The respondents‟ income level moderates the relationship between service quality
and customer satisfaction
66
Research Question 5:
Which demographic variable, if any, moderates the relationship between customer
satisfaction and customer loyalty?
Hypotheses to address RQ 5
H1: The respondents‟ age group moderates the relationship between customer
satisfaction and customer loyalty
H2: The respondents‟ gender group moderates the relationship between customer
satisfaction and customer loyalty
H3: The respondents‟ occupation moderates the relationship between customer
satisfaction and customer loyalty
H4: The respondents‟ education level moderates the relationship between customer
satisfaction and customer loyalty
H5: The respondents‟ income level moderates the relationship between customer
satisfaction and customer loyalty
*** Please note that the hypotheses must appear from research question.
From hypothesis statement, the researcher will determine the variables required
in the study, how these variables will be measured, method of data collection,
and the appropriate statistical analysis employed to test the hypothesis. That is
how the hypothesis provides a clear direction for the research.
67
Chapter II - Review of Literatures Related To the Study
68
Customer Loyalty is defined as a customer who would continue to purchase
goods or services from the same company whenever possible, and who
continues to maintain positive attitudes towards goods and services from the
company
69
2.4 The Theoretical Framework of the Study
The theoretical framework for your study is the schematic diagram drawn to
indicate how you believe the variables involved in your study should relate to
each other. This is important since it will become the heart of your study.
Service
Quality
Customer Customer
Satisfaction Loyalty
Corporate
Image
70
Firstly, the service quality of a university is hypothesized to influence
students‟ loyalty towards their university. Mathematically, the equation is:
Y1i = β0 + β1X1i + εi.................................................(i)
In this section, you should describe clearly your research plan. The
audience wants to know your plan of work and the activities necessary for the
completion of your project. To keep the audience with you, begin with a few
attractive introductory paragraphs that restate the research problem, research
objectives, and your research questions.
71
3.1 The Target Population
Target population is the element in general population that you want to cover
in your study. For example, if your study wants to cover hardcore poor in the
community, then your population should be the people who earn less than
RM475 per month. You need to define your target population precisely and
explain the rationale for selecting that specific population for your study. Make
sure you choose the right population; otherwise, your conclusion might be
questionable since you might be shooting the wrong target.
The target population for this study is the undergraduate students in the
Faculty of Business and Management at Universiti Teknologi MARA (UiTM).
UiTM has branch campuses in all states in the country. Since these branch
campuses are widely scattered in term of geographical location, the study
applied the cluster sampling method whereby the peninsular Malaysia is
divided into five geographical regions namely Eastern Region, Northern
Region, Southern Region, Western Region, and East Malaysia.
72
characteristics in term of location, environment, facilities, and climate condition,
the study could select one or more than one branch campus to represent the
respective region.
Explain in detail how you plan to draw a representative sample from the
population. Since the data from this sample will be used to draw conclusion on
the overall population, tell your audience how you ensure the selection process is
representative and unbiased. What are your sampling method, and the rationale
for using that method? Convince the audience of your knowledge regarding
statistical validity and reliability. Tell specifically the sampling process you plan to
execute. The process include sampling frame, how you obtain the sample, how
you ensure the sample is randomly selected, how you contact the respondents,
the instruments to use for data collection, who does the data collection for you,
are they qualified for the job, how do you ensure the data collection will take
place accordingly without prejudice, bias, or cheatings etc.
Once the branch campus is selected, the list of students in the faculty of business
and management is obtained from the office of academic affairs at the respective
branch. The study obtained only the list of students in semester 3 and above
since the students who have been in the university longer are in a better position
to provide their opinion for items under the respective construct of the study.
Below are the steps taken in the sampling and data collection processes:
1. The permission to collect data is obtained from the faculty of business and
management at the respective campus. At the same time, the list of students
to be used as a sampling frame for the study is also obtained from the faculty
together with the class schedule where the students are having their lecture.
73
2. The respondents are selected from the sampling frame using the random
numbers generated through the computer. For example, if the sampling frame
consists of 100 students and the study decided to take only 20 students as
respondents, then the 20 numbers are generated using the following
computer statement (Rand 20 1-100). The computer would generate
randomly 20 numbers in the range of 1 to 100. Let‟s say the 20 generated
numbers are 5, 9, 11, 17, 23, 25, 36, 39, 44, 47, 51, 58, 66, 68, 73, 78, 88, 89,
96, and 98. Then these numbers are matched with the list in the sampling
frame. The students whose number on the list matched the numbers
generated by the computer would become the respondents for the study.
3. Once the respondents are identified, the research assistants would proceed
to the respective lecture room during the time where the students are having
their lecture session. The research assistants would obtain the permission
from the respective lecturer to distribute the questionnaires to the respective
respondents.
4. If the research assistants failed to find the respondents in that particular time,
they would proceed to the next lecture session or with the help of their friends,
the questionnaire is delivered to the respective respondent.
5. The respondent can respond to the questionnaire at their convenient time and
mail the completed questionnaire using the self-addressed envelope. In order
to ensure confidentiality of the responses, each questionnaire is enclosed with
a cover letter that explains clearly the purpose of the study. The respondents
do not need to provide any identification on the questionnaire. They could
also submit the completed questionnaire using the sealed envelope to the
representative at the office of academic affairs the following day.
The study employed the above method in order to allow respondents to attend
the questionnaires without any pressure so that whatever response provided
reflects the true opinion from the respondents themselves. Therefore, the
response bias due to time constraint and the presence of researcher would not
arise.
74
3.3 The Measuring Instruments employed in the Study
All processes and procedures must follow certain assumptions to avoid the
threat of invalid methodology. For example, if the sampling distribution of your
variables does not meet the required assumptions for statistical distribution, then
the statistical methods used in the data analysis is invalid. Hence, all conclusion
based on statistical test is considered to be invalid.
75
4. -Respondents answer the questionnaire at their own convenience without
bias, prejudice, threat, influence or pressure from any party.
Describe the procedure you plan for data analysis. Explain how you will
treat the questionnaires – the checking, editing, coding, and data transcribing.
Describe how you will address each individual research question. What
variables are you looking for? How these variables provide answers to your
research questions? What statistical computations are required? State the types
of statistical test you plan to perform. This is important when you have certain
hypotheses to prove.
The researcher analyzed the data using the Statistical Package for Social
Science version 18 (SPSS 18.0) and AMOS program version 18.0 (AMOS 18).
This SPSS helps to facilitate data clearing, and checking for logical
inconsistencies. The researcher carried out data mining procedure before the
actual data analysis is performed. In order to make your proposal more
convincing, particularly your proposed hypothesis; you should include the
relevant statistical analysis procedure that you want to employ in testing your
proposed hypotheses.
76
Example 14a: The Hypothesis and its Statistical Analysis
All research studies have limitation and a finite scope. Since the results of your
study are not valid for the whole world, not applicable across all industries, not
apply to all population, and might not be true under all circumstances, you
should think carefully about the scope and limitations of your study. It is
irritating to you when someone tells you that your research is not valid under
certain conditions, which is within your limitation!
77
Example 15: The scope of the study
Like in other studies, this particular study has its own limitations, both in scope
and methodology. First of all, the study was carried out in the Faculty of
Business and Management. Thus, the result might only be generalizable to the
above population. In other words, the findings might be different if the scope is
increased to include more faculties since different faculties might pose
different characteristics.
78
EXERCISES
1. You have been approached by the chief editor of “People Today” to carry
out a research study. The magazine has been unsuccessful in attracting
corporate officers to purchase the magazine. As a research consultant, you
need to understand and gain insight on the problem of the study. You
believe that an exploratory design will help them to identify the problem that
is causing the resistance among the corporate customers. You are also
required to prepare a research proposal for the marketing manager.
b) Describe the exploratory technique that you would use in defining the
problem of the study and explain how it is done. Is it appropriate for you
to proceed using the descriptive design? Why?
c) State two research questions that could help you to identify and solve
the chief editor‟s problem.
2. Describe how you would develop the theoretical framework for your study.
3. State and describe three important variables that are needed in developing a
theoretical framework. Illustrate using a schematic diagram.
79
Chapter 4: Sampling Design
The main reasons for studying sample instead of the whole population:
1. Cost constraint: Working with a sample does not incur much expenses
compared to working with the whole population.
2. Time constraint: Working with sample does not require longer time
compared to working with the whole population.
Elements
Elements are the selected respondents or objects from the defined target
population from which the researcher will study by collecting the required data
and information.
Sampling Units
Sampling Frame
Sampling frame is the list of all eligible sampling units and elements in the
population from where the sample can be selected. Examples of sampling
frame are a phone directory, a list of companies selling health products, a list of
diabetic patients, a list of voters in the area, a list of houses in the district, a list
of credit card holders who apply for additional credit limit, etc. Using the
appropriate sampling technique the researcher can select the sample from this
sampling frame for the study.
The choice of sampling design to be employed in the study depends very much
on certain factors deemed to be crucial. Among the factors that should be
considered are:
Accuracy is the degree to which bias is absent from the sample. There is no
variation in measures due to any influences that causes skewness in the
distribution with an accurate sample. The degree of accuracy required or the
researcher’s tolerance for error may vary from one project to another. This is
true especially when cost constraint or other considerations may be traded
off for a reduction in accuracy.
3. Availability of Resources
If the researcher’s financial and human resources are restricted, these
limitations will most certainly eliminate the more sophisticated time-
consuming and complex probability sampling methods to be employed.
Marketing practitioners who are concerned by the cost versus the precision
of the information often choose the non-probability sampling method for their
study rather than do not conduct any study at all. This is because most of
marketing research projects is concerning the customers’ opinion, customers’
preference, and customers’ perception about certain product or service in the
market; and these types of researches do not require the measurement
which is too precise like the scientific research.
4. Time Constraint
Researchers who need to complete a project in a short time will be more
likely to select a simple, less time-consuming sampling method rather than a
more complex and accurate method. For instance, a telecommunication
company would like to introduce a new plan, which requires them to do a
research on that matter. If the study is too time-consuming, several
prospective customers might move away from the company to their
competitors. So, doing the research within the time frame, the company will
not only keep current customers coming back but also encourage
prospective customers to buy the new plan.
2. Sampling error can be determined only after the sample is drawn and data
collection has been completed.
From the principles of the Central Limit Theorem (CLT), the degree of
sampling error and its impact can be reduced by increasing the size of sample to
be taken from its population. CLT is the theorem which states that for any target
population, the sampling distribution of the mean and proportion derived from a
random sample will be approximately normally distributed provided that the
sample size is sufficiently large. The CLT plays an important role in
understanding the concepts of sampling error, the statistical significance, and the
determination of sample size.
1. By using the appropriate sampling technique when drawing the sample from
its population.
2. By increasing the size of sample or respondents to be obtained from its
population.
2. The level of confidence desired in the estimate (CL). The higher the level of
confidence desired in the findings, the larger the sample size needed.
Confidence interval represents a statistical range of values within which the
true value of the target population parameter is expected to lie.
The non-sampling errors are biases that may occur in any research study
regardless of whether a sample or a census is used. These errors may occur at
any stage of the research process.
Sampling Design
Probability Non-probability
Probability sampling methods are the most commonly used method because the
selection of participants is determined at random. In the probability sampling, any
element of the population has an equal chance or equal probability of being
selected as a sample. Thus the representativeness of a sample cannot be
questioned.
Probability
Proportionate Disproportionate
The process of simple random sampling consists of the following four steps:
b. The list of all elements in the population. This list is sometimes called a
sampling frame form which the sample will be selected. For example, a
complete list of marketing students in UiTM is a sampling frame.
d) The respondents selected this way are totally random; hence the
problem of biasness does not arise.
b. The sampling fails to use all the information of a population that might
result a waste of the design.
Like simple random sampling technique, the systematic sampling technique is also
applicable for the homogenous population. In this sampling technique, the element
which falls on the kth number on the list is chosen as a respondent. This sampling
procedure fixed the increment in order to determine the following respondent, for
example, every 5th person on the list is selected as respondents.
a. Selection process
iii Determine the interval (k) when k is the ratio between the size of
population and size of sample (k = N/n). For example if N = 500 and
n = 25. In this case, K = 500/25 = 20.
5 25 45 65 85
105 125 145 165 185
205 225 245 265 285
305 325 345 365 385
405 425 445 465 485
When the characteristic of the population under study is not homogenous, then the
researcher has to stratify them into homogenous groups. When the characteristic in
each group is homogenous, then the researcher can employ either the simple
random sampling or systematic sampling technique to randomly select the
respondents. The stratified sampling is the process of selecting a sample that
represents each stratum in a population. The researcher can stratify the
heterogeneous population into homogeneous population within a stratum in terms of
socio-economic status, ethnicity, religion, gender, marital status, type of houses, etc.
a. Specify the strata (elements within the same strata is homogenous, and
elements between strata is heterogeneous).
c. Now we can select the sample within each strata using simple random
sampling since the population is already uniform.
Sample from
Whole Population in each stratum
Population each stratum
Single (10)
Total Single (100)
investors
who bought Married (400) Married (40)
the ASB unit
trust in Jan Divorced (300)
2008 is 1000 Divorced (30)
Widowed (200)
Total = 1000
Widowed (20)
Total = 1000
Total = 100
The advantage of stratified sampling is the estimate obtained from the sample
is more accurate since it ensures all strata are represented in the sample.
The researcher also can compare whether different strata perform differently
regarding certain variable of interest.
Cluster sampling is the probability sampling technique applicable when the study
covers a large geographical area, the population is not homogeneous, and the
sampling frame is not available. The clusters can be the area in a residential district
(section 1, section 2, etc), the growth corridor in the country (WPI, NCER, ECER,
Prof Dr Zainudin Awang 92
UniSZA
Chapter 4: Sampling Design
etc), or the developmental region such as eastern region, northern region, central
region, and southern region.
The process of cluster sampling is:
a. First of all, the researcher needs to specify the clusters of interest within
the population. The researcher can classify certain groups into a cluster in
one geographical area.
b. For example, UiTM has campuses scattered across the country. The spread of
UiTM campuses can be classified based on the respective region namely eastern
region (Kelantan, Terengganu, Pahang), northern region (Perlis, Kedah, Penang,
Perak), central region (Shah Alam, Petaling Jaya, Meru), southern region (Johor,
Melaka, Negeri Sembilan), and east Malaysia (Sabah, Sarawak). In this case,
each region is a cluster and under each cluster there are a few campuses
available.
c. In the study, firstly the researcher needs to select randomly a few clusters
from the available clusters (regions). Secondly, the researcher needs to
select randomly a few campuses among the campuses available in the
selected clusters.
d. All students in the selected campuses are taken as respondents in the
study. In this case the sampling technique is applied in the selection of
clusters and in the selection of campuses within the selected clusters.
Thus, no sampling technique is applied on the respondents in the selected
campuses since all of them will be covered in the study.
In the non-probability sampling, the element in the population does not have equal
chance of being selected as a sample. This technique is used in research where no
inference concerning the population needs to be made. This means the result of the
study is only applicable to the particular sample and does not necessarily reflect the
whole population. However, this type of sampling is very economical and easily
performed.
Non-probability
Convenience sampling offers the least expensive and least time consuming
compared to all sampling techniques. In addition, the sampling units are accessible,
easy to measure, and cooperative. Often, it is used in exploratory research for
generating ideas on certain phenomena, insights into certain problem, to develop
hypotheses, to pre-test questionnaires, and to conduct pilot studies.
c. The sampling process is not appropriate for population inferences, and the
technique is not recommended for descriptive and causal research.
Prof Dr Zainudin Awang 94
UniSZA
Chapter 4: Sampling Design
Example 4.1:
Suppose a firm wishes to investigate consumers who own digital versatile disc
(DVD) players in their home. Since different brands of DVD pose different
characteristics, the researcher may wish to ensure that each brand of DVD is
included in the sample. In a quota sampling, the researcher determines the number
of respondents (quota) to obtain for each DVD brands. For example, a researcher
may choose 100 respondents to interview. The respondents must consist of
customers using the following DVD brand, 40 Sony, 30 Samsung, 20 Toshiba, and
10 of other brands.
a) The defined target population is very small and the characteristic of the
population is unique.
For example, the research project is to study the success story of single
mother in the direct-selling business. First of all, the researcher needs to find
the first respondent, the single mother doing direct-selling business. After
obtaining the required data (through personal interview or structured
questionnaire) from this respondent, the researcher would ask that
respondent to locate her friends who possess the same characteristics. The
researcher would stop once he has obtained enough number of respondents
required for the study.
Exercises
True-False Questions
STATEMENT TRUE FALSE
A list containing elements in a target population is known as a
1.
sampling frame.
A sample is any subgroup of a population selected for
2.
participation in a study.
A stratified sampling is usually the most efficient sampling
3.
technique in a statistical sense.
A sampling unit is the object in the population from which
4.
information is desired.
Resource constraints should not be used in determining sample
5.
sizes for research project.
6. Systematic sampling does not require a sampling frame.
7. Quota sampling does not guarantee representativeness.
A census involve the complete enumeration of all of the
8.
elements of a population.
The sampling system consists of a list of directions for
9.
identifying the target population.
A census can increase non-sampling error to the point that
10.
these errors exceed the sampling errors in a sample.
a. Population
b. Census
c. Sample
d. Sampling unit
a. Quota
b. Sample
c. Census
d. Panel
a. Sample
b. Quota
c. Population
d. Census
7. Among the factors that the researcher should consider when determining the
sample size for his research is:
8. A good sample is the one in which there is no bias from the sampling process.
This is defined as _____________.
a. Precision
b. Consistency
c. Accuracy
d. Random
a. Group
b. Bi-variant population
c. Sample
d. Market
a. Sampling frame
b. Population
c. Sample size
d. Sampling unit
11. When the sampling error cannot be calculated due to the method used to
select the sample, the researcher has just used a (n) _______________.
12. A researcher begins by selecting one customer from the list of 1300
customers, and then continues the process by selecting every 30th customer
on the list is using a ______________.
a. Judgement sampling
b. Quota sampling
c. Proportional stratified sampling
d. Snowball sampling
a. Quota sample
b. Convenience sample
c. Judgement sample
d. Snowball sample
15. What type of sampling is suitable when there are great variations among the
sub-groups in the population?
a. Stratified sampling
b. Cluster sampling
c. Systematic sampling
d. Simple random sampling
16. Which of the following techniques is typically the least expensive and least
time consuming?
a. Simple random
b. Convenience
c. Stratified
d. Cluster
a. Systematic sampling
b. Proportional stratified sampling
c. Disproportionate stratified sampling
d. Quota sampling
Essay Question
b. What is the probability sampling method that is appropriate for the above
opinion poll? State your reason(s).
c. Assuming that the list of 2800 third year students is obtained from the
registrar office, the numerical list of names is between 1 and 2800. Using
the table of random numbers below, explain the procedure you would
follow to select six samples from the list.
2. a. Give two main reasons for using cluster sampling over simple random
sampling.
b. What are the main reasons for using a sample rather than a census?
c. The 5 year old Shah Alam housing directory lists the residents’ names
alphabetically together with their addresses. TNB wishes to conduct a
research on the perception of Shah Alam’s residents towards the
campaign on using Solar Panel for house use. For the above situation:
a. Convenience sampling
b. Judgmental sampling
c. Snowball sampling
c) What are the classifications of non-sampling errors that could arise from
the study?
10. A prominent bank located in Kuala Lumpur has some 400,000 users of its
credit cards scattered throughout Malaysia. The application forms for the
credit card asked for the usual information about name, address, income,
level of education and so on that is typical of such applications. The bank is
currently interested in determining if there is any directional relationship
between the frequency of card usage and the amount of credit charged on the
card for various purchases made within a month, i.e. the more frequent the
card is used, the higher is the credit amount charged on the card.
a) Identify the sampling frame that would be used by the bank to conduct the
investigation. How is the sampling frame acquired?
b) Choose a suitable sampling technique and indicate how you would draw a
sample from the target population using the sampling technique chosen.
c) Determine and describe the appropriate data collecting method for the
study. State TWO (2) advantages and disadvantages for choosing the
method.
Primary data source is the source that presents first-hand information. The
sources can be the original records, original documents, personal
interviews, and records of eyewitnesses that are original in nature. In other
words, primary sources are the original data collected by the researcher
himself for his study. Primary data sources can come from direct
outcomes of an event the actual experience that is recorded by the person
involved. Among the examples of primary data collection activities are:
1. The review and analysis of the existing data on the target markets. These
data are available in the business magazines, research studies,
government publications, annual report of a company.
4. The analysis of voting trend among the Malay voters towards the non
Malay candidates in the 2008 General Election.
The secondary research does not require any method to get the data
because it already exists. Thus, no specific method or technique is needed to
collect information. Instead, the time and effort are spent to locate and gather
the information from reliable sources. Some resources for secondary
research information include:
Sometime the researcher does not have to work with numerical data in his
research project. For example, the study on public‟s perceptions concerning
certain government policy, the study on customers‟ acceptance towards
Islamic Insurances, or the public preference concerning new Proton models
just launched in the market. The above research project is called qualitative
research. Qualitative research provides definite market information regarding
the opinions, perceptions, and behaviours of the subjects in the market
research study. The information obtained is in terms of non-numerical.
Qualitative data is less accurate and limited statistical methods available for
data analysis. Qualitative are measured according to their specific categories.
For example, gender is measured as „male‟ or „female‟. Marital status is
measured as „single‟ or „married‟, opinion is measured as „strongly disagree‟,
„disagree‟, „neutral‟, „agree‟, or „strongly agree‟ and so on. Qualitative
research is employed to achieve a variety of objectives. Among the objectives
of qualitative research are:
Personal interviews and focus group discussions are the most common
qualitative research techniques.
a. Personal Interviews
When the numerical data is required in the study, the research is called
quantitative research. Data on demand, supply, interest rate, annual income,
total export, total import, and the volume of daily sales are numerical. The
quantitative research creates statistically valid market information. The
information is in terms of number. For example: The weight of new Nokia
hand phone is:
1. Personal Interview
b. Too time consuming, not suitable for big project with small budget.
2. Telephone Interview
j. Some people are not keen to talk long over the phone.
3. Mailed Questionnaire
4. Self-Administered Questionnaire
a. Lower budget required since the questionnaire is self explained and can
be distributed to many respondents at the same time.
5. Direct Observation
a. Researcher can record the real scenario since the respondents do not
know that their attitudes are being observed.
c. Not suitable for serious type of research project where the actual
responses from the respondents are required.
Prof Dr Zainudin Awang 112
UniSZA
Chapter 5: Types of Data, Measurement, and Scales
Nominal scale data describes the variable in terms of its category and the
measure differs in terms of quality rather than quantity. The nominal scales
are assigned to represent the categories of which the characteristic of data
belongs. Nominal scale measure is the most basic measure among the four
levels of scale designs and used mainly for the purpose of identifying the
respondents in terms of their gender, ethnic groups, socio-economic status,
preferences, etc. In this scale, the question requires respondents to select
only one descriptor as their response. The responses provided do not contain
any level of intensity; therefore, it is impossible to establish any form of rank
order or to calculate any meaning values to describe the respondent.
Prof Dr Zainudin Awang 113
UniSZA
Chapter 5: Types of Data, Measurement, and Scales
b) Gender of respondent
Male
Female
Example 5.2
Instruction
Rank the following brands of shampoo based on your preference.
Example 5.2a
Instruction
Please indicate how much you agree or disagree with the following
statements concerning the entertainment program on TV. Select the
most appropriate number to represent your opinion.
1. Strongly Disagree
2. Disagree
3. Not Sure
4. Agree
5. Strongly Agree
This method is the most commonly used in business and social sciences to
obtain data concerning respondent‟s attitude, opinion, perception, or intention
on something. The method produces an interval-scaled data since the data
fall anywhere between two extreme intervals namely “strongly Disagree” and
“Strongly Agree”
Please indicate how strongly you agree or disagree with the following
statement concerning the corporate image of this university. You can mark
your sincere response anywhere between 1 to 10. Please remember, there is
no right or wrong answers!
The ratio scale describes variables that have equal intervals between them
but also have an absolute zero. In the simplest terms, this means they are
variables for which one possible value is zero or the actual absence of the
variable or trait is possible. Ratio scales assign absolute rather those relative
quantities. For example, both monetary values and weight are ratio scales
because they possess absolute zeros and interval properties. The absolute
zero represents a point on the scale at which there is an absence of the given
attribute. Ratio scale is the highest level of measurement among the four
scales and the statistics that can be used in the analysis are the most flexible.
However, in behavioural research interval scale typically is the appropriate
measurements.
Table 5.2: The Relationship between Nominal, Ordinal, Interval, and Ratio
Nominal Ordinal Interval Ratio
scale Scale Scale Scale
Preference
Preference Money spent in last
Name of ratings of food
No. rankings among two months at the
restaurant quality in the
the restaurant respective restaurant
scale of (1 to 10)
1. Delima Café 2 5.5 $180.00
2. Sigai Café 1 7.1 $300.00
Example 5.6: The illustration of real-life application for four different scales.
The marketing graduate plans to open a fast-food restaurant near the campus.
She conducted a study on the students‟ preference towards the existing
restaurants in the area. Her sample consists of five main restaurants in the
area.
Using the nominal scale, she assigns the five restaurants as follows:
1 = Delima Café, 2 = Sigai Cafe, 3 = Selera Kampung, 4 = Cenderawasih, and
5 = Restoran MakDara. In this case, she uses the numbers to nominate the
restaurants. Thus, in the above numeric, it does not imply that Delima Café,
which was assigned 1, was in any way the most preferred restaurant as
compared to Restoran MakDara, which was assigned the number 5. The
number only represents the name of the respective restaurant, and not the
performance of that particular restaurant compared to other restaurants.
Nominal scale allows the researcher to run some descriptive statistical analysis
such as percentages and mode. However, the statistical analysis such as mean
is not meaningful in this level of measurement.
Using the ordinal scale, the respondents were asked to rank the five
restaurants based on their order of preference. Rank 1 indicates certain
restaurant is the most preferred restaurant. Based on the result in Table 5.2,
restaurant Sigai is the most preferred while restaurant Cenderawasih is the
least preferred restaurant in the area.
By using the interval scale, the respondent‟s preferences for each restaurant
can be expressed on a 10 point rating scale. The respondents can provide their
score from 1 to 10 to reflect how strong their preference towards each particular
restaurant. From the table, Cenderawasih restaurant obtains rating 1.0 while
MakDara restaurant obtains rating 4.0. However, the rating for MakDara 4.0
compared to Cenderawasih 1.0 does not, in any way, imply that the customers
prefer MakDara restaurant 4 times when compared to Cenderawasih.
The ratio scale in Table 5.2 shows that a respondent was asked to indicate the
amount of money he spent at each of the four restaurants in the last two
months. Based on the table, the respondent spent $100 in MakDara and only
$10 in Selera Kampung. Here we can say this particular respondent spent 10
times as much in MakDara as Selera Kampung. The $0 score indicates that the
particular respondent did not spend his money at that restaurant.
There are many ways to classify the scales of an object in the study.
However, most of the researcher employs three types of scale to measure the
response namely categorical scale (sometimes-called voting scale), comparative
scale (sometimes-called ranking scale), and rating scale.
Voting scale has several response categories and used to elicit responses
concerning the objects, events, or respondents. Basically, voting scale is used to
judge properties of objects without reference to other similar objects. The voting
scale may be in the forms of two categories “like or dislike” or other
classifications using more categories such as three-point, five-point, seven-point
or more points. One argument is that more points on a scale provide greater
sensitivity of measurement.
1. Dichotomous Scale
The dichotomous scale is used to elicit one of two responses (Yes or No)
and sometimes three responses as in the example below. Basically, a
nominal scale is used to vote the response. This method produces a
nominal scaled data.
Example 5.7
Do you like pizza?
Yes ____
No ____
Not Sure ____
2. Category Scale
The category scale uses multiple items of object characteristics to elicit a
single response. It is often used to separate the objects according to their
categories (gender, age, etc.) This method also produces the nominal
scaled data.
Example 5.8
Please state your gender?
Male ____
Female ____
3. Likert Scale
Likert scale was developed by Rensis Likert and designed to examine how
strongly the subjects (respondents) agree or disagree with the given
statements about the stimulus object. Typically, each scale item has five
specific responses ranging from “Strongly Disagree” to “Strongly Agree”.
The example below illustrates how likert scale method is used in
measuring the image of hospital.
The respondents are asked to indicate their attitude about the object being
rated. The likert scale applied below would produce the ordinal scaled
data since each number correspond exactly to one response statement.
Hence, the data can only be analyzed using the non-parametric methods,
not the parametric methods.
Example 5.9
Instructions
The following statements are about the image feature of this hospital. Please indicate the
extent to which you think this hospital possess the feature described by each statement by
ticking at the appropriate number of the scale provided on the top.
Example 5.10
Instructions
Please marked (X) in the blank which best indicates what you think about your
organization. This university is:
5. Numerical Scale
The numerical scale is quite similar to the semantic differential scale with
bipolar labels placed at the endpoints. The difference is the five points or
seven-point numerical scale in between the two endpoints as shown in the
example below. This method produces an interval scaled data. Thus, the
researcher could employ the parametric as well as the non-parametric
methods to analyze this data.
Example 5.11
How comfortable are you with the new arrangement in the classroom?
6. Stapel Scale
Stapel scale is developed by Jan Stapel. This scale is designed to
simultaneously measure the attitude of respondents at both direction
(positive and negative), using 6 (+3 to -3) or 10 categories (+5 to -5), without
a neutral point (zero). This scale obtains an interval data since it does not
have an absolute zero point.
Example 5.12
Instructions
Please circle the number to evaluate how accurately each word or phrase below
describes the healthcare services in this hospital. The larger the plus number
you mark the more accurate the phrase describes the services. The minus
number shows the less accurate that you think the phrase describe the service.
Example 5.13
Instructions
Please tell us how much you like this class by pointing to the face that
best describes your feeling.
1 2 3 4 5
In ranking methods, the subject directly compares two or more objects and
makes a choice among them. It is often used in the situations where the
respondent is asked to select one as the “best” or the “most”. This method is not
satisfactory when more than two choices are given. For example, when the
respondent is asked to select the most preferred among three brands A, B, and
C. Assuming that 40 percent choose brand A, 30 percent choose brand B, and
30 percent choose brand C. The question is, “Which is the preferred brand?” The
answer suggests that A is the most preferred is correct, but 60 percent of the
respondents choose the same brands other than A. The researcher can avoid
the confusion by using the following types of comparative or ranking scale.
Example 5.14
Instructions
Example 5.15
Instructions
Rank the following brands of digital camera that you would like to buy in
order of preference. Begin by assigning 1 for the most preferred and 5 for
the least preferred.
Example 5.16
Instructions
You are given 100 points to allocate to five attributes of mobile phone.
Please allocate your points to indicate the importance of the attribute in
a mobile phone.
1. Design
2. Colour
3. Shape
4. Size
5. Price
Total points 100
4. Comparative Scale
The comparative scale provides a benchmark or a point of reference to
assess attitudes toward the object, or situation under study. The example
below illustrates the use of comparative scale in assessing service quality in
the university.
Example 5.17
Instructions
Over the last two years, the overall quality of my university service has:
1 2 3 4 5 6 7
Decline Remained Improved
Significantly the same significantly
Prof Dr Zainudin Awang 126
UniSZA
Chapter 5: Types of Data, Measurement, and Scales
Exercises
True-False Questions
STATEMENT TRUE FALSE
The brands or types of stores are classified or measured using
1.
a nominal scale.
The numbers in nominal scale reflect the amount of the
2.
characteristic each object possesses.
An ordinal scale reflects the magnitude of differences between
3.
objects.
Data analysis using Regression methods requires at least an
4.
interval-scaled data.
Ratio-scaled measurements assign actual quantity rather than
5.
those relative quantities.
6. Whenever you divide people into groups you are probably using a
___________ scale of measurement.
a. Ratio
b. Ordinal
c. Interval
d. Nominal
8. If we use a scale using 5 categories of races, what type of scale would this
represent?
a. Nominal
b. Ratio
c. Interval
d. Ordinal
9. For which of the following types of scales is the mode the most appropriate
measure of central tendency?
a. Ordinal
b. Ratio
c. Interval
d. Nominal
Essay Questions
b) Classify the variables “service quality”, “class status”, and “parents income”
asked in the questionnaire below into quantitative or qualitative variables.
Then state whether they are nominal, ordinal or ratio level of measurement
3. State the levels of measurement (nominal, ordinal, interval or ratio) for each
of the following variables of interest:
4. In the following situation, decide whether you would use a personal interview,
telephone survey, or self-administered questionnaire. Give your reasons.
b) A survey on the residents in Shah Alam on why they choose the area.
You also wish to get some information about what they like and do not
like about this area.
5. You plan to conduct a mail survey of the lecturers in the whole of UiTM
across the country. The study concerns the Universities policies regarding the
payment of moving expenses of employees who are transferred. Suggest 4
ways to improve the response rate of such a survey.
6. a) The following questions are not written clearly. Improve on the questions.
i. What is your income?
RM 900 and less
RM 901 to RM 1000
RM 1001 to RM 2000
RM 2001 to RM 3000
RM 3001 to RM 4000
RM 4000 and more.
i. Age : 34 years
ii. Marital Status : Married
iii. Annual Family Income : RM 100,000
b. Identify the highest and the lowest levels of measurement and state their
differences.
8. Design a question and name the level of measurement you would use for the
following:
b) To find out students‟ preference among the three types of exams namely multiple
choices, essay type and a combination of both.
9. Over the past five years, a manager has noted a steady growth of sales and profits
for her division‟s product line. Is there a need to utilize any secondary date for further
evaluate the division‟s condition. Give your reasons.
10. What would be the most appropriate method of data collection under the following
situations? Give reasons for your answers.
11. Indicate whether the following measures are nominal, ordinal, interval, or ratio
scales:
a) A student survey:
How frequent do you get to campus using each of the form of
transportation available per week? Select your answer based on the
following scale:
1. Most frequent
2. Frequent
3. Never
Prof Dr Zainudin Awang 132
UniSZA
Chapter 5: Types of Data, Measurement, and Scales
4. Seldom
5. Very seldom
i. Walk …………….
ii. Bicycle …………..
iii. Public transportation …………..
iv. Motor vehicle …………..
b) A survey of employers:
Should the company continue its generous medical insurance
program?
Yes No
13. Indicate whether the following examples are nominal, ordinal, interval or ratio
scales. Explain briefly the meaning of these scales.
15. For each of the following, indicate the type of measurement scale it
represents, that is nominal, ordinal, interval or ratio scale.
i. Weight of an automobile
ii. Exterior colour of an automobile
iii. Size classification of an automobile
iv. Original cost of an automobile
v. Rankings of automobile models by a consumer organization
vi. Trunk size of an automobile
16. a) A researcher wishes to compare between two counters (HEA and HEP) on
the following attributes: friendly personnel and convenience of location.
17. a) Distinguish between rating and ranking scale. Which is a better attitude
measurement?
A: Profile
Primary data is the data obtained by the researcher or the agent hired by the
researcher. The persons can be the research assistant, the staff of consultant
firms, or the individuals who are doing the data collection work on behalf of the
researcher.
On the other hand, secondary data is the data already obtained and compiled by
other parties such as government departments, company‟s annual report,
newspapers, business magazines, or the internet web pages. This type of data is
already published by relevant parties for their own purpose such as the annual
economic reports prepared by Bank Negara to the government, the company‟s
annual report prepared by the company to its shareholders, the productivity
report produced by the department of statistics, and the annual report prepared
by each ministry.
A researcher can obtain data from secondary sources for his research but
these data might not be suitable and accurate for his research since these data
might be outdated, or the data collected by the previous parties were for different
objectives and had used different scales. Hence, whenever possible, the
researcher should collect his own data unless the required data is not possible to
obtain from primary sources.
For example, the data regarding the monthly volume of palm oil production
in Malaysia, the daily price of gold and petroleum traded on The World Market,
the dividend payments made by ASB to the investors, and many other data are
called historical data and not available through primary sources. These data are
only accessible from the Department of Statistics, the Bank Negara report, the
Ministry of Finance, the Ministry of Primary Industries, the Malaysian Commodity
Exchange etc. In other words, the secondary data are data obtained from the
secondary sources
Once the researcher has determined his research objective, research question,
and the corresponding hypothesis for his research project, what he needs to do
now is to collect the required data. Data is information from sample that the
researcher would analyse in order to meet his research objective, facilitate his
research question, and prove his research hypothesis. For example, the data or
information about customers are gender, age, qualification, marital status,
number of kids, monthly income, brand of car, type of house, religion, hobby,
sports, leisure activities, credit cards, golf membership etc.
Direct
Personal Self-Administered Observation
Interview Questionnaires
Mechanical
Face-to-Face Mail Survey Observation
Interview
Computerised Content
Telephone Questionnaires Analysis
Interview
1. Survey Method
2. Structured Questionnaires
One of the most popular methods of collecting research data is through the
structured questionnaire. These questionnaires are self-explained and self-
administered. In using this method, the researcher designs carefully a series
of questions that cover the variables of interest in the study such as the
respondents‟ demographic background, their opinions concerning certain
issues, their perception concerning certain service performance, and their
intentions to do in the future etc. Structured refers to the degree of
standardization imposed in the process of data collection (questionnaires). In
other words, the researcher arranges the questions properly on a paper
together with the cover letter to explain the purpose of data collection, the
instruction to the respondents on how to respond to the questions, and the
assurance of confidentiality of information provided.
Advantages of a Questionnaire
c. More respondents and wider area can be covered. The researchers can
send his questionnaires to as many respondents as he likes since the cost
for each respondent is very small.
Disadvantages of a Questionnaire
a. The poor rate of response from respondents. Since the researcher does
not have gestures or face-to-face contact with respondents, the
respondents can just throw away the questionnaires. In this method, the
researcher should make a regular follow up contact with respondents
using letter or telephone call. Usually the researcher would call the
respondents earlier to inform them that he is sending a questionnaire to
obtain data for his specific objective. The call is made as a gesture in
order to obtain good cooperation from respondents.
Private agents who conduct surveys found that people are more likely to
respond to mail questionnaire that has professional appearance and come
together with attractive prizes such as lucky draws for the respondents who
send the completed questionnaire in time. Some questionnaires obtain good
responses from respondents especially the questionnaire that comes together
with the warranty cards when customers purchase certain products.
Customers will have to answer series of questions on the warranty cards
before sending back to the manufacturer for product warranty.
Personal Interview
1. Face to Face Interview
In the face-to-face interview (sometimes called personal interview), the
researcher will prepare the questions to be asked during the interview with
respondents. Each question represents the variable that the researcher
wants to obtain its data. The questionnaires cover all variables required from
a respondent. Before the interview begins, the researcher would explain the
objective of research, ask for cooperation, and give assurance that the
Prof Dr Zainudin Awang 139
UniSZA
Chapter 6: Primary Data Collection Methods
response given is only for research purpose and the information is treated
with confidential. This is important since personal data is confidential, and no
one will reveal his personal information if the confidentiality is not assured.
The interviewer should posses good personality so that the interview session
proceeds smoothly and in a friendly atmosphere. During the interview, the
interviewer will read the question and record the response. Personal
interviews may be conducted at the respondents‟ home, offices, or
anywhere. Below are the common examples of places which are basically
used to conduct the interview.
a. Door-to-door Interview
a. Higher response rate. With proper plan and approach, the respondents
will not turn down the request for an interview. The interviewer must be
friendly and creative in getting cooperation from respondents. The
interview method normally achieves a response rate of more than 70%.
d. People will usually respond with good gesture and provide accurate
responses when approached in person. The Malay culture of not saying
“no” still holds true when someone comes to the door with polite and
peaceful manner.
2. Telephone Interview
c. Easy monitoring. The researcher can monitor the interviewers‟ work more
easily since he can check the telephone numbers of respondents and the
time called.
a. Lower response rate. The rate of response is lower since the respondents
can just “hang-up” the call when he realised it is time consuming, or
unsure of the confidentiality of the conversation.
3. Self-Administered Questionnaire
In this section, we will discuss another survey method that does not require
the presence of the interviewer. In this technique, the researcher distributes
questionnaires to respondents personally, through mail service, by inserting
in the newspapers, or by sending to the email address. The difference
between self-administered questionnaires compared to personal interview is
in the self-administered questionnaire, the respondents make an effort to
read and respond to the questions. However, the effectiveness of self-
Prof Dr Zainudin Awang 143
UniSZA
Chapter 6: Primary Data Collection Methods
4. Mail Survey
5. Mail Panel
6. Computerised Questionnaire
a. Internet Survey
Lately, the internet poll has become one of the popular methods to obtain
information from the public especially their opinion concerning certain
issue of public interest. In internet survey, the researcher brings an issue
to attention and requests the opinion from the public. The public can
respond by voting the statement that closely resembles their opinion. At
the same time they can view the current standing in term of the most
popular opinion and the corresponding vote obtained.
Internet survey appears when a computer user is asked to go to a
particular Web site location and answer the series of questions displayed
in the website. In this technique, the respondents are not selected using
specific technique but those who visit the Web site where the survey is
posted are invited to participate in the survey.
b. E-mail Survey
1. Low cost.
2. Very high data collection speed.
3. Non interviewer bias
1. Direct Observation
In this method, the researcher will identify his respondents and records the
required data based on what he observes. This method is suitable for a
research to study the behaviour of respondents. For example, the research is
carried out to identify how car drivers behave on the road during traffic jam.
In his observation, the researcher will record the type of drivers who follow
the traffic rules properly, and the type of drivers who choose to ignore rules
such as queue jumping or overtaking using emergency lanes. The
researcher may also be interested to know the types of vehicles
(motorcycles, cars, buses, and lorries) which always ignore traffic rules
during traffic jams. Another area where this method is suitable is to observe
customers‟ behaviour in the supermarket. In the market research study for
example, the researcher is interested to know how customers make
decisions in choosing which shampoo to buy. The researcher will stand in the
area where hundreds of shampoos of different brands are placed on the
rack. He will record how the customers choose the shampoo. Most probably
some customers have decided earlier which brand to buy; some customers
will compare prices, packaging, and even the smells before buying the
shampoo. The researcher will record specific characteristics of customers
who prefer certain brands of shampoo etc.
1. The respondent will not act his normal behaviour if he knows that he is
being observed.
2. Mechanical Observation
2. Less flexible.
3. Content Analysis
a. High degree of disguise since the data is collected over the situation to
be observed.
Exercises
True-False Questions
STATEMENT TRUE FALSE
Primary data is the data obtained by the researcher or a person hired
1.
by the researcher to collect data.
Method for collecting primary data consists of survey method,
2.
observation method and quantitative data.
In survey method, questionnaires are given to respondents to elicit
3.
information for the study.
Structured questionnaire refers to the process of drafting questions
4.
such as question content, question wording, and question ordering.
5. Telephone interview is expensive than face-to-face interview.
The effectiveness of self-administered questionnaire will depend on
6.
the efficiency of the written words rather than skill of interviewer.
A mail panel consists of a large, nationally representative sample of
7.
households that have agreed to participate in mail questionnaires.
The computerised questionnaire consists of internet survey, email
8.
survey and CD-ROM databases.
Content analysis obtains data by observing and analyzing the
9. contents or messages of advertisements, newspaper articles,
television programs, etc.
Mechanical observation requires the respondents‟ direct participation
10.
in the study.
a. Conclusive research
b. Scientific observation
c. An experiment
d. A survey
a. Telephone survey
b. Personal interview
c. Mall intercept interview
d. Mail survey
a. Personal interview
b. Telephone interview
c. Mail survey
d. Internet survey
8. Which of the following is the best reason for using telephone interviews?
a. Accuracy of information
b. A two way relationship exists
c. A high response rate
d. The cost is higher compared to other methods.
a. Be patient
b. Pay attention to accuracy and details
c. Keep responses confidential
d. All of the above
Essay Question
b) What type of survey method would you choose when you are dealing in
the following situations?
i. Survey the achievement of industrial engineers
ii. Survey the satisfaction of rent-a-car users.
iii. Survey the awareness of television advertisements
iv. Survey the mission of corporate executives.
2. a) A telephone interviewer asks: „I would like to ask you about ethnic groups
in East Malaysia. Are you an Iban, Bajau or Dayak?‟ After the respondent
replies, the interviewer says, „We have conducted a large number of
surveys with people of your background and we do not need to question
you further. Thank you for your cooperation‟. What type of sampling
technique was utilized?
4) What factors would you consider when choosing between observation and
survey methods?
CHAPTER
7 THE SECONDARY DATA SOURCES
Secondary data is the data collected by other parties. This data might be
obtained for a different purpose. The current researcher who uses secondary
data has little knowledge about the original respondents, the sampling
methods, the method of data collection, and the accuracy of the measuring
instruments employed by the previous researcher to obtain data. Thus, the
researcher needs to be careful concerning its validity before using the
secondary data.
Secondary data utilized to serve three purposes for the existing research project.
First, it serves as a specific reference of research in some point in time. For
example, the total population of Malaysia in 1990 was 25 million; the volume of
Malaysian cars on the road in 2000 was 500 thousand units etc. Second, it is
used to support a larger research study. Research procedures typically called for
at least a minimum amount of exploratory study to learn if the past trend can
contribute to the present study. Secondary data will help to avoid the researcher
from “reinventing the wheel” and provide the relevant reference for developing
hypothesis. Finally, secondary data may be used as the sole basis for a research
study, especially research that requires the use of historical data such as time
series analysis etc
The main advantage of secondary data comes from the availability and easily
accessibility. This is particularly true when electronic retrieval such as internet is
used to access digitally stored data. However, the researcher should be cautious
when using the secondary data because these data have some limitations and
disadvantages. The limitation of secondary data is that they were not meant for
the research at hand. The usage of secondary data to the current research might
not be relevant and accurate because of the following reasons:
1. Internal Sources
2. External Sources
External sources are more varied than the internal source and the
researcher should always find better methods for the search. External data
are created, recorded, or generated by an entity other than researcher’s
organization. Published data can be classified into five categories. First
category is computerized database that includes of both online and CD-
ROM databases. Access to these databases is available through the online
service system.
In searching for secondary data, the researcher normally faces two difficulties,
namely:
After the researcher has made some preliminary search into encyclopedias
and textbooks, he might have some knowledge about the topic. This
knowledge will help to develop a bibliography for the study which is the
essential step for secondary data search and much important to the study.
The term “CD-ROM” stands for Computer Disk Read Only Memory.
Data is stored on CD-ROM disks and read from the CD-ROM through
computer screen or printed on a printer. The disks are purchased on a
subscription basis from a publisher or vendor. Basically, the libraries
Prof Dr Zainudin Awang 158
UniSZA
Chapter 7: Secondary Data Sources
subscribe to this database for the search process and update the
information. The example of this database is Infotrac (Information
Access Company) which is widely used and contained three indexes;
General Periodicals, Legaltrac and National Newspaper Index.
Doing a database search requires the researcher to think through the topic
very carefully and to plan the research strategy. Understanding the specific area
can help to clarify the topic and highlight possible problem areas. However,
several steps might be useful in conducting computerized search as the
following:
a. Speed; the computer can scan many citations in a short amount of time.
a. Imprecision; the computerized search allows only one or more key words
typed into the system. As a result, synonyms and alternate spellings are
not automatically searched.
b. Wide capability; the computerized search will often find many sources that
contain the key words that the researcher has used in the system, but the
citation is not related to the subject of research.
In searching the information from this source, the researcher must obtain
information about the title of the books. This information can be obtained
from the general bibliography of books in the library. The researcher
should make sure that the key words about the topic used is not too
narrow and if needed should expand into different categories to ensure
that searching process is smooth. Basically, the researcher obtains
secondary data from books for the purpose of supporting his ideas and to
illustrate something such as the underlying theory of the research project
at hand.
Actually, periodicals are the best single source of information for the
researcher because there are thousands of publications covering almost
all subjects. Periodicals refer to professional journals, public opinion
surveys, business periodicals, etc., which contain important information
to support research. Professional journals comprised of Journal of
Management, Journal of Marketing, Journal of Education, Journal of
After the researcher has gone through all the three sources, several
sources might be useful in searching secondary data for a research
project. There are many other sources such as data published by
business organization, non-governmental organization, local
governments, etc. One of the major miscellaneous sources is the
publications from higher education such as university’s bulletin,
academic researches, and doctoral dissertations. For example, UiTM
publishes its bulletin for internal circulation called Buletin UiTM every
three months; UiTM also published the research papers produced by its
academic staff every year after the annual Conference on Scientific and
Social Research (CSSR).
The researcher should ask the following questions to evaluate secondary data:
However, the researcher may also look at the following criteria, issues, and
remarks for the purpose of obtaining suitable secondary data for the study.
Exercises
True-False Question
STATEMENT TRUE FALSE
Primary data are data which are collected for purpose other than
1.
the current study at hand.
The question of its relevance and accuracy are common
2.
disadvantages of using secondary data.
Secondary data should be carefully examined in order to
3.
determine how key variables are defined and measured.
Relevance and accuracy are common advantages of using
4.
primary data.
Before collecting external secondary data, internal secondary
5.
data sources should be carefully examined.
Published external sources of secondary data are broadly
6.
classified as general business data or government data.
Full-text databases are composed of citations to articles in
7.
journals, magazines, and newspapers.
Dependability of data is typically determined through examining
8.
the credibility or reputation of the source.
The accuracy of secondary data is about the same for most
9.
countries in the world.
The researcher is ethically obliged to ensure the relevance and
10.
usefulness of secondary data to the problem at hand.
1. Data collected for some purposes other than the problem at hand is known as:
a. Primary data
b. Secondary data
c. Qualitative data
d. Quantitative data
4. The data originated by a researcher for the specific purpose of addressing the
problem at hand is called ___________.
a. External data
b. Internal data
c. Primary data
d. Secondary data
6. If you want to examine the income statements and balance sheets of one of
your company’s biggest competitors, you would use a ____________
database.
a. Geographic
b. Citizen
c. Financial
d. Social
Prof Dr Zainudin Awang 165
UniSZA
Chapter 7: Secondary Data Sources
7. The starting point for any search for secondary data should be __________
a. Internal sources
b. Original sources
c. Acquired sources
d. External sources
WHAT IS QUESTIONNAIRE?
167
Chapter 8: Designing a Questionnaire
The steps involve in designing questionnaire can be divided into 4 stages. The
first and second steps are included in the first stage called the identifying
process. The second stage consists of drafting questions. It consists of three
steps namely,
168
Chapter 8: Designing a Questionnaire
The third stage is organizing the questions, deciding the form and its
layout. Finally, after the questionnaire has been developed, a series of test
should be conducted to eliminate sources of error and assuring the accuracy of
data (see Figure 8.1).
Identify what information is required for this research. There are few steps
involved:
1. Identify the research questions concerning the problem which the top
management must address.
169
Chapter 8: Designing a Questionnaire
The researcher should identify and determine the data collection method to be
used in the study before designing the questionnaires. This step involves the
following tasks:
a) Open-ended question
170
Chapter 8: Designing a Questionnaire
b) Close-ended question
171
Chapter 8: Designing a Questionnaire
Example 8.1
Instruction:
I would like to assess how strongly you agree or disagree with every statement
provided here concerning your university. You are free to indicate your
response between one (1) to ten (10). Your score of (1) would indicate you
strongly disagree with the statement and score (10) would indicate you strongly
agree with the respective statement.
Remember, you can tick your response anywhere between 1 to 10.
The statement provided How strongly you agree or disagree with the statement
172
Chapter 8: Designing a Questionnaire
What are the items you purchased from this store in the last two days?
173
Chapter 8: Designing a Questionnaire
d. The question content should cover the issues and obtain sufficient
data for specific problem. The researcher should avoid double-
barrelled questions that attempt to cover two issues in a single
question. In the following example, the researcher wants to cover
two issues (variety and taste) in a single question.
Indicate how far you agree or disagree with the following statements.
The foods provided at the dining hall are varied and tasty.
The response from the respondents will not be accurate since there are two
issues in one question. Furthermore, the respondents might refuse to
answer if they agree with one issue but not with the other issue. It is
advisable to break this question into two questions as shown in the
following example.
Indicate how far do you agree or disagree with the following statements.
The food provided at dining hall has variety.
The food provided at dining hall is tasty.
174
Chapter 8: Designing a Questionnaire
However, the following guidelines might help the researcher in the process
of wording question.
175
Chapter 8: Designing a Questionnaire
Another problem is the use of technical jargon. The jargons will confuse
respondents and they might refuse to respond since the terms used are
quite alien to them.
Do you think that the market of shoes in this area is already saturated?
176
Chapter 8: Designing a Questionnaire
If you want to make certain assumption, you should ensure that the
assumption is worded explicitly in the question to prevent the respondent
from overestimating the main subject or topic of question.
177
Chapter 8: Designing a Questionnaire
Do you like swimming? (The response might be „yes, but there is hardly a
swimming pool in this area‟ or „no since the swimming pool is not available
in this area)
i. Do you think Proton Savvy is the best car in its class since it is proven in
the test conducted by an expert?
ii. Cigarette smoking has high probability of male impotence and many
cases of sexual dysfunction; my question to you is “Do you smoke?”
178
Chapter 8: Designing a Questionnaire
Do you think the government should reduce the tax in order to reduce the
smuggling activities of products into the country?
Do you think the government should reduce the tax in order to reduce
smuggling activities of products into the country, or you think that
government should determine other mechanism to reduce the smuggling
activities?
After the researcher has properly worded the questions, the next process is
to determine the response options for each question. There are two options
that range from the open-ended response (free choice of words) to the
closed-ended response (specific alternatives provided). Open responses
refer to the freedom given to the respondents to phrase their answer. Close
responses typically are categorized as dichotomous or multiple choice. The
best decision in designing the response is by giving the respondents some
freedom to express their opinions. This can be achieved by providing more
choices of responses and ask the respondents to indicate the most
appropriate response.
179
Chapter 8: Designing a Questionnaire
a. Ordering by section
b. Opening question
The opening question is extremely important since it will establish
rapport, build confidence, and encourage respondents to complete the
questionnaire. This question should be simple, interesting and not
threatening.
180
Chapter 8: Designing a Questionnaire
The respondent is less likely to ignore the question after rapport has been
established and the respondent has been involved in answering general
and specific question. Therefore, a difficult or sensitive question must be
placed last in each section in the questionnaire. For example, the question
on income and telephone number must be the final item in the section of
classification question.
Example 8.14:
Strongly Strongly
Disagree Agree
1 2 3 4 5 6 7 8 9 10
Once the researcher has drafted the questions, he would organize them in
a form or layout called questionnaires. The forms of a layout such as the
format, instructions, spacing, and positioning of questions in the self-
181
Chapter 8: Designing a Questionnaire
v The questions should be ordered from light to heavy tone and also
from general to specific.
182
Chapter 8: Designing a Questionnaire
iii The questions should not be crowded into a page in order to show
the questionnaire look simple. It will cause the questionnaire to look
complex which will result in lower cooperation and completion
rates.
v A time estimate is given so they know when to return it. For face-
to-face interview this is not needed.
183
Chapter 8: Designing a Questionnaire
vii It begins with the request for a sincere participation and ends with
sincere thanks for cooperation.
Dear respondent:
I am doing an academic research entitled:
Congratulations, you have been selected as a respondent for this study. Your
sincere cooperation is crucial in determining the success of this study. In this
study, I would like to assess your perception concerning certain factors of your
university. Please read carefully all instructions pertaining to every section and
answer every question sincerely. There is no right or wrong answer. All
responses provided will be treated with strict confidential and will be used for
this academic research only. Thank you very much for your cooperation.
Sincerely,
………………......................………
Assoc Prof Dr Zainudin Hj Awang
Centre for Graduate Studies (CGS)
Universiti Teknologi MARA (UiTM)
15100 Kota Bharu Kelantan
e-mail: zainudin888@kelantan.edu.my
184
Chapter 8: Designing a Questionnaire
In debriefing technique, the researcher does not inform the respondents about
the pre-testing. The respondents are told about the pre-testing process only
after they have completed the questionnaire. The probing of questions is done
through asking respondents to explain their response for each question in the
questionnaire. The respondents are also encouraged to state any problem they
encountered while answering the questionnaire. The weakness of this
technique is the difficulty to get cooperation from respondents. Hence, the
number of successful probing is limited.
185
Chapter 8: Designing a Questionnaire
Dear respondent:
Congratulations, you have been selected as a respondent for this study. Your
sincere cooperation is crucial in determining the success of this study. In this
study, I would like to assess your perception concerning certain factors of your
university. Please read carefully all instructions pertaining to every section and
answer every question sincerely. There is no right or wrong answer.
All responses provided will be treated with strict confidential and will be used for
this academic research only.
Thank you very much for your cooperation.
Sincerely,
………………………………….........
Assoc Prof Dr Zainudin Hj Awang
Centre for Graduate Studies (CGS)
Universiti Teknologi MARA (UiTM)
15100 Kota Bharu Kelantan
e-mail: zainudin888@kelantan.edu.my
186
Chapter 8: Designing a Questionnaire
Instruction:
I would like to assess how strongly you agree or disagree with every
statement provided here concerning your university. You are free to indicate
your response between one (1) to ten (10). Your score of (1) would indicate you
strongly disagree with the statement and score (10) would indicate you strongly
agree with the respective statement.
Example:
The statement provided How strongly you agree or disagree with the
statement
187
Chapter 8: Designing a Questionnaire
Please indicate how strongly you agree or disagree with the following
statement concerning the corporate image of this university.
You can mark your sincere response anywhere between 1 to 10. Please
remember, there is no right or wrong answers.
188
Chapter 8: Designing a Questionnaire
Please indicate how strongly you agree or disagree with the following
statement concerning the quality of services at this university.
You can mark your sincere response anywhere between 1 to 10. Please
remember, there is no right or wrong answers.
189
Chapter 8: Designing a Questionnaire
190
Chapter 8: Designing a Questionnaire
Please indicate how strongly you agree or disagree with the following
statement concerning your satisfaction being a student of this university.
191
Chapter 8: Designing a Questionnaire
Please indicate how strongly you agree or disagree with the following statement
concerning your loyalty towards this university.
You can mark your sincere response anywhere between 1 to 10. Please
remember, there is no right or wrong answers.
192
Chapter 8: Questionnaire Design
1 Your Gender:
Male ………................................ 1
Female …………............................. 2
2 Your Age
Below 20 years …………............................. 1
20 - 22 years …………............................. 2
23 - 25 years …………............................. 3
Above 25 years …………............................. 4
Thank you very much for your sincere cooperation in this study.
SEasonS
HOTEL
K U A L A L U M P U R
Dear Guest,
We are honoured that you have selected our Hotel for your recent business trip/leisure
travel. Thank you for your patronage.
The objective of the Legend Group of Hotels and Resorts is to provide our guests with
gracious service and comfort. We would very much appreciate your help in achieving this
objective.
We randomly select a number of guests and invite them to let us know their feelings
about their current hotel stay. As one of our valued guests, we seek your co-operation in
completing this questionnaire and returning it by folding this form and securing the
adhesive flaps.
All information provided will be used in our planning to serve you better in the future.
Thank you for your kind assistance. We look forward to your next visit with us.
Yours sincerely,
1. Taking all things into consideration, how would you rate our overall services and
facilities? From a scale of 1 being extremely dissatisfied to 10 being outstanding,
how best would you describe your experience?
1 2 3 4 5 6 7 8 9 10
□ □ □ □ □ □ □ □ □ □
2. Please tick the boxes that best describe your feelings about the items listed below
with which you have had experience this visit.
1 2 3 4 5 6 7 8 9 10
Overall Staff Courtesy
Overall Staff Helpfulness
Overall Staff Efficiency
Door Service
Luggage Service
Front Desk Service
Housekeeping/Maid Service
Room Comfort
Efficiency of Bathroom Equipment
Value for Money
Please comment
________________________________________________________________________
________________________________________________________________________
________________________________________________________________________
3. Below are some of the other services which you may or may not have used. Please
tick the boxes that best describe your feelings about these services which you may
have used on this visit.
1 2 3 4 5 6 7 8 9 10
Laundry/Valet Service
Telephone Service
Business Centre Service
Message Service
Recreation Facilities
Food & Beverage
Room Service
Overall Restaurant Service
Overall Lounge and Bar Service
Room Maintenance
Please comment
________________________________________________________________________
________________________________________________________________________
________________________________________________________________________
4. Below are the individual Food and Beverage outlets, Banqueting and other facilities
which you may or may not have used. Please share your experience with us by rating
their service and facilities
Coffee House
Seasons Café 1 2 3 4 5 6 7 8 9 10
Food Quality
Service
Friendliness
Lounge
Lobby Lounge 1 2 3 4 5 6 7 8 9 10
Food Quality
Service
Friendliness
O’las Restaurant 1 2 3 4 5 6 7 8 9 10
Food Quality
Service
Friendliness
Room Service 1 2 3 4 5 6 7 8 9 10
Food Quality
Service
Friendliness
Timeliness
Bar
Kasbah 1 2 3 4 5 6 7 8 9 10
Food Quality
Service
Friendliness
Entertainment
Food Quality
Service
Friendliness
Banquet / Conference 1 2 3 4 5 6 7 8 9 10
Set-Up
Food Quality
Service
Friendliness
Helpfulness
Light and Sound
Please comment
________________________________________________________________________
________________________________________________________________________
________________________________________________________________________
6. Thinking of your most recent previous stay, how would you rate us in terms of
overall services and facilities on that particular experience?
1 2 3 4 5 6 7 8 9 10
□ □ □ □ □ □ □ □ □ □
8. Based on what you have had experienced, would you recommend this Hotel to your
friends and business associates?
□ Yes □ No
9. If you were to return to this city, would you stay in this Hotel again?
□ Yes □ No
10. Did you experience any problem which is of concern to you during your stay with
us?
□ Yes □ No
11. If so, in which of the following areas and what was the problem?
Please specify
______________________________________________________________
Comment
______________________________________________________________
□ Yes □ No
14. Compared to other hotels which you have had experienced, how would you rate our
Hotel?
15. Which Hotel would be your next preferred choice to stay in aside from this Hotel?
□ Business □ Leisure
□ Conference/Corporate Meeting □ Airline Stopover
18. Please tick below your country/state of residence and write down your country/state
of origin on the space provided.
Country/State of origin
__________________________________________________
Malaysian States
□ Africa □ Sabah
□ Australia □ Sarawak
□ Brunei □ Selangor
□ Canada □ Wilayah Persekutuan – Kuala Lumpur
□ China □ Wilayah Persekutuan – Labuan
□ Europe □ Wilayah Persekutuan – Putrajaya
□ Hong Kong □ Penang
Once again, thank you for returning this questionnaire. We look forward to
welcoming you back to Grand Seasons Hotel.
Name: __________________________________________
Room No: _________
Address: ________________________________________
E-mail: _________________________________________
Today’s Date: ________
Exercises
True-False Questions
a. Translate the information needed into a set of specific questions that the
respondents can and will answer.
b. Uplift, motivate, and encourage the respondent to become involved and
cooperate with the interview.
c. Minimize response error.
d. Accomplish all of the above.
a. Hypotheses.
b. Research questions.
c. Problem definition.
d. Both a and b are true.
a. Make certain that the form and layout of the questionnaire is appropriate.
b. Prepare a set of dummy tables to show how the data analysis will be
structured.
c. Carefully decide on question order.
d. Reproduce the questionnaire to make sure it is readable and useful.
a. Age group
b. Education
c. Occupation
d. All of the above characteristics.
a. Structured questions
b. Unstructured questions
c. Disguised questions
d. Undisguised questions.
8. A structured question with two response alternatives, such as yes and no, is
called a:
a. Double-barrelled question.
b. Dichotomous question.
c. Telescoping question.
d. Leading question.
11. Open ended questions that respondents answer in their own words are
known as:
a. Unstructured questions.
b. Structured questions.
c. Dichotomous questions.
d. Telescoping questions.
a. Categorical variable
b. Unknown variable
c. Dependent variable
d. Continuous variable
13. What type of question is the following: “I‟m going to show you a sample of
magazine advertisement. What does this advertisement mean to you”?
a. Open-ended
b. Closed-ended
c. Dichotomous
d. Multiple choice
14. “Smoking is not just a bad habit but it is the main cause of cancer. I would
like to ask you one question, Do you smoke?” is an example of:
a. a generalization question
b. an implicit alternative question
c. a position bias question
d. a leading question
a. Early
b. In the middle
c. Late
d. Anywhere
19. Which form of response is considered superior when you wish to discover
opinions and degrees of knowledge?
a. Open-ended questions
b. Dichotomous questions
c. Multiple-choice questions
d. None of the above
20. Which of the following are reasons for NOT including a question in
questionnaire?
Essay Questions
1 The President of UiTM Alumni Association wishes to carry out a survey from
the members who were from classes of 2010 onwards to determine their
achievements, current activities and future aspirations. Information pertaining
to the following areas is desired: gender of members, major area of study,
grade point average, further educational pursuits (that is master‟s degree or
doctorate), current annual salary, number of full time position held since
graduation, annual salary anticipated in five years, political party affiliation
and marital status. You are asked to conduct the survey.
2. The management of UiTM is wondering why there are many students who
refuse to consume the food that is provided at dining halls at every college.
Every day, a large amount of food is not taken and wasted, whereas the
government is paying the contractor certain amount of money per head for
the food. This has resulted in a waste of millions of ringgit every year. In
addition, every day a large amount of food is sold cheaply to local farmers to
feed their livestock. Instead of eating the food provided at the dining halls,
these students buy food at the cafeteria or from the stalls nearby.
The management of UiTM wants you to look into this matter and carry out a
research. The findings of your research should determine the main factors
that contribute to the above problem so that appropriate actions can be taken
to rectify the situation.
At the same time, you should come out with the recommendations (based on
your findings) so that the government fund is not wasted, and the students
can use their monetary loan to buy books or save for their future.
i. Nominal (1 question)
ii. Ordinal (1 question)
iii. Interval (1 question)
iv. Ratio (2 question)
Assoc Prof Dr Zainudin Awang 205
UiTMKelantan
Chapter 8: Questionnaire Design
a. Double-barrelled question
b. Loaded question
c. Leading question
d. Ambiguous question
4. a) The following questions are not written clearly. Rewrite the questions.
o Never
o Occasionally
o Always
o Family
o Partner
o Friends
o Others
i. Quantitative data
ii. Qualitative data
5. State five most common mistakes that should be avoided when developing a
questionnaire. Give examples of each.
6. What are the factors that constitute bad questions in questionnaire design?
Develop TWO (2) examples each of bad and good questions
CHAPTER
9 DATA MANAGEMENT
The goal of any research is to obtain the required data, analyze these data
accordingly, and write a report based on the output of data analysis. Raw data
is simply a recorded measure of certain information collected to accomplish the
objectives of the study. The conversion of raw data into a piece of information
using appropriate statistical analysis for answering the stated research
questions is called data analysis.
After data have been collected from the respondent, the next step is to analyze
them using various software programs such as SPSS, SAS, Excel, etc. However,
before we begin to analyze the data, we need to conduct some preliminary steps
to ensure that the data are ready for further analysis (see Figure 9.1).
Data Editing
Data Coding
Error checking
and verification
Data Entry
Data Analysis
Interpretation of output
1. Data Editing
The purpose of data editing is to increase the accuracy and precision of the
data by checking certain omissions, such as blank response on the
questionnaire, legibility of handwriting, inconsistent responses, etc. The
researcher can check several areas of concern namely:
the “good service” into a more specific response such as “provide fast,
efficient, and friendly service”.
There are a few options which the researcher can take when facing with the
problem of “no response” or “missing data”. Among those options are:
The researcher can assign a missing value using a dot (.) in a data entry. The
computer recognizes a dot as no response for that particular item.
iv. The proportion of missing data for each respondent is large and it is
impossible to edit.
v. The missing data involved is the main variables in the study.
Nevertheless, the decision to exclude the respondents from the study may be
subjective and might lead to bias results. If the researcher has to resort to this
procedure, he should record the procedure adopted to identify these
respondents and their reference. A major problem in editing is subjectivity.
There are many procedures to be adopted by the researcher in doing editing
process. A good researcher should have a systematic procedure for assessing
the questionnaires.
2. Data Coding
After the editing and cleaning process of raw data is completed, these data
have to be coded in order to facilitate data entry. If the data is not coded
systematically, it will cause the lack of clarity of data, which will result later in
serious confusion in the analysis. Data coding refers to the process of
identifying and classifying each response with a numerical score or symbol. In
other words, the researcher needs to assign a code (usually number) to each
possible response in a question. Generally, codes are numbers or symbols.
However, codes are broadly defined as rules for interpreting, classifying, and
recording the data. In this process, the researcher must specify the column
position, number of variable, name of variable and coding instructions.
Coding will facilitate the process of transcribing the data from the
questionnaire into the computer. In the coding process, the researcher will
look through every question and each response to prevent confusions
especially when there are many questions in a questionnaire and a large
number of respondents in the study. The codes are numerical because
numbers are easy to input into a data file. For example, using a scale from 1 to
10, how would you rate the following statement concerning your university?
In the above example, the code of response is readily available for each item.
2) What is the main reason you prefer this fast food restaurant?
Tastes (1)
Cleanliness (2)
Promotion of the restaurant is interesting (3)
Others, please specify_________________________ (4)
For an open-ended question, it is difficult to code because the responses may vary
largely among respondents. The researcher should develop codes for the
anticipated responses. It begins by generating a list of as many potential responses
as possible. These responses can be given values within a range determined by the
actual number of separate responses identified.
Example 9.1
The example extracted from the research study done by Zainudin (2004).
Demographic Information:
1. My Gender :
Male …………………………………….. 1
Female …………………………………….. 2
2. My age:
Below 25 years ………………………………...….. 1
25 – 35 years …………………………………….. 2
36 – 45 years …………………………………….. 3
46 – 55 years …………………………………….. 4
Above 56 years ……………………………………... 5
3. Image Scale
To what extent would you agree or disagree with the following statements
concerning the services provided by this hospital? Circle 1 if you strongly
disagree, and circle 7 if you strongly agree with the statement. You can also
circle anywhere in between depending how strongly you agree or disagree with
the given statement.
Out-patient services
In-patient services
Second step:
After you have reviewed all the factors involved, you must sort out the factors such
as the name of variable, number of question, variable label and value of responses
which are important in coding process. Below are some illustration on how to sort
out the important factors in the question 3 and 4.
Variable Name
Customer Contacts Strongly Strongly
Disagree Agree
3. The service delivered by staffs/doctors is within my 1 2 3 4 5 6 7
expectation.
4. The doctors/staffs of this hospital treat their patients 1 2 3 4 5 6 7
respectfully.
Third step:
Next, write down all the factors you have sorted out in a sheet of paper according
to their category. Table 9.1 below shows the examples of how to write all the
factors according to their codes. The column number, types of data and scale
measures are shown in the columns 2, 4 and 7. Refer to the spreadsheet resulted
from the “variable view” in the SPSS 16.0 software.
Final step:
After you have completed furnishing the table above, it is possible to key in data
directly from the questionnaire. However, you must properly check the coding
table to ensure the accuracy of the coding especially on the variable name,
variable label, and value of responses before data entry can begin.
Example 9.2
The following are a few responses obtained using the open-ended questions about
frozen food.
All of these answers can be categorized under the factor of “dislike frozen
foods” and assign the code 1. The construction of category reflects the judgment
of the researcher. A major objective in the coding process is to accurately transfer
the meanings from written responses to numeric codes. Therefore, some
researchers develop and assign codes based on their experience or theoretical
consideration prior to the beginning of the fieldwork. The following are several
guidelines that the researcher might use for coding open-ended questions.
1. Differentiate Coding
Differentiating the response categories for open-ended questions is more
difficult than fixed alternative questions. The category of codes must be
mutually exclusive. Generally, the researcher performs a test tabulation in
order to identify the verbatim responses and then establish the coding
categories that reflect the judgment of the respondents. Test tabulation is a
test on a small sample of respondent that respond to a particular question to
identify the stability and distribution of responses that will determine the coding
scheme.
2. Confirm Coding
In this stage, the researcher should assign categories to use in answering the
research question although there is no respondents mention about them. For
example, the management of UiTM is concerned about the preferred menu of
food in a research study of food services at dining hall. However, there is no
respondent mentioned about the preferred menu on the open-ended question
concerning food services at dining hall. The researcher should assign this
critical issue in order to know that no one has mentioned a particular
response.
3. Revising Coding
After test tabulation has been conducted, the researcher might have some
useful categories in mind for answering research questions. However, the
difficulties will arise if the preliminary tabulation contained too many codes, or
the codes are too abstract, or there are only a few categories established. The
codes should be revised and reduced into smaller categories to answer the
research question.
In revising the codes, the data should be coded as detail as possible. For
example, in a study on counter services in UiTM Library, the term
“Responsible Staff” might consist of friendly staff, polite staff, helpful staff,
courteous staff, and cute staff. However, the detailed code provides too many
categories. A good researcher would generalize the heading pertinent to the
research questions.
THE CODEBOOK
1. Column number
2. Record number
3. Variable number
4. Variable name
5. Question number
6. Instructions for coding
3. Data Entry
Data entry includes tasks involved with the direct input of the coded data into
some specified software package that ultimately allows the research to
manipulate and transform the raw data into useful information. Normally, the
software packages employed are SPSS package and Microsoft Excel. For
example, the SPSS Data Editor comprises of spreadsheet in which you can
enter, edit, and view the contents of the data file. Each row of the editor
represents a case, and each column represents a variable.
One critical task of data entry personnel is to ensure that the data entered are
correct and mistakes free. How to detect the errors in data entry? Error
detection normally occurs after the data have been entered into computer
storage. Entry errors can be detected with error edit routines built or developed
into the data entry software. An additional approach is to visually scan the actual
data after they have been entered. A frequency data table is one of the
approaches for visually scanning entered data. All missing values should appear
with a period (dot) in the cell. It is possible to add, change, or delete values
easily after the data have been entered. Once the missing values and the
recodes are taken care of, the data are ready for analysis.
Exercises
True-False Questions
a. Data editing
b. Sample selection
c. Data coding
d. Recording the response
a. Record; code
b. Code; field
c. Field; record
d. Code; recode
a. Verbatim response
b. A summary of responses
c. Validation codes
d. Single column codes
a. Project code
b. Codebook
c. Fixed field code
d. Validation code
10. ___________ includes the tasks of direct input of the coded data into computer.
a. Data coding
b. Data editing
c. Data entry
d. Data transformation
Essay Question
3. What is missing data, and how does it affect the research study?
7. Discuss the importance of codebook and state the information contained in the
codebook.
9. The data were collected via in-home personal interviews and 1200
questionnaires were returned from the field. During the process of editing and
cleaning the data, you found that 39 questionnaires have some missing
responses.
After data entry has been completed, the researcher needs to carry out data
analysis in order to achieve his research objectives, address his research
questions, and to test his research hypotheses. First of all, he has to perform
data mining for all measuring items in the study. Secondly, he has to perform
the descriptive analysis for important demographic variables to verify that
respondents are represented across their demographic characteristics. Thirdly,
he has to run the required analysis to address his research questions. Finally,
he has to perform the statistical analysis in order to test the stated hypotheses
concerning the study.
This step is necessary in order to screen the data from “outliers”. This is an
important procedure to ensure that the data collected is normally distributed. The
proof of normality for the data is extremely necessary in order to meet the
requirement for employing the parametric statistical test. Otherwise the researcher
needs to employ the non-parametric statistical procedures. A common “mistake”
committed by novice researchers is running the parametric statistical analysis
without proving that the data at hand meet the required assumption for the
respective analysis to be valid. Caution should be adhered that certain statistical
analysis is not valid if the data collected do not follow normal distribution. In this
case, the proof of normality for data at hand is required before a researcher could
proceed further.
Example 10.1
We will carry out data mining process using the data obtained by Zainudin (2007)
in his study on University Education. In this example, we select one variable
namely the corporate image of a university for further analysis. In his study,
Zainudin (2007) employed 16 items to measure the corporate image of a
university.
Table 10.1 illustrates the descriptive scores for each of the 16 items based on the
responses from 872 respondents. The table also presents the standard error,
standard deviation, and the measure of skewness for every item in the study.
From Table 10.1, the measure of skewness for all items fall between -0.971 and -
0.316. The general statistical measure of skewness ranges from -3.0 and 3.0.
The measure of skewness close to -3.0 indicates the data is distributed extremely
skewed to the left, while the skewness close to 3.0 indicates the data is distributed
extremely skewed to the right. If the data is perfectly normally distributed, the
measure of skewness is 0.0. However, the measure between -1.0 and 1.0 is
considered normally distributed and acceptable to proceed with the parametric
analysis procedure.
Otherwise, if the data is not normally distributed, the researcher has to employ
non-parametric procedure for his statistical analysis. Since the measures of
skewness for all items are within the range between -1.0 and 1.0, the researcher
can conclude that the distribution of data does not depart from normality. Hence,
the required assumption for employing the parametric statistical procedures for
further analysis is satisfied.
cont...
Example 10.2
Using the same dataset, the researcher analyzes the demographic variables
namely the respondents’ gender, age group, and their duration at the university.
Tables 10.2 to 10.8 show how the respondents are distributed according to the
selected demographic characteristics.
Valid Cumulative
Gender of respondents Frequency Percent Percent Percent
Male 404 46.3 46.3 46.3
Valid Female 468 53.7 53.7 100.0
Total 872 100.0 100.0
The above table shows that the respondents are represented almost equally in
both gender categories namely male and female.
Valid Cumulative
Age of respondents Frequency Percent Percent Percent
20 - 22 Years 595 68.2 68.2 68.2
23 - 25 Years 277 31.8 31.8 100.0
Total 872 100.0 100.0
The above table shows that the respondents sampled only fall into two age
categories. Meaning, in this study, the respondents aged below 20 years and
above 25 years are not covered. Hence, the generalization from this study is only
valid for the students’ population aged between 20 to 25 years.
cont…
Prof Dr Zainudin Awang 226
UniSZA
Chapter 10: Data Analysis
The above table shows that the respondents are taken almost equally across three
durations of study at the university. However, the students less than two
semesters in campus are not included. Hence, the generalization is not applicable
for students who are less than two semesters being in the university.
Cumulative
Parent’s income (RM) Frequency Percent Percent
Less than RM 1000 261 29.9 29.9
RM 1001 - RM 3000 222 25.5 55.4
RM 3001 - RM 5000 177 20.3 75.7
RM 5001 - RM 7000 109 12.5 88.2
More than RM 7000 103 11.8 100.0
Total 872 100.0
The above table shows that the respondents are taken across all income levels
which reflect the socioeconomic status of their parents. Meaning the generalization
from this study is valid for all income categories of students’ population.
cont…
The above table shows that the respondents are taken across all education levels
which reflect the socioeconomic status of their parents. Again, the generalization
from this study is valid for all education levels of students’ population.
The above table shows that the respondents are taken across all five education
levels which would also reflect the socioeconomic status of their parents. Since
respondents from all occupation levels are sampled, this study is generalizable to
respondents from all levels of occupation.
cont…
The above table shows that the respondents from both choice categories namely
“first choice” and “not first choice” are taken in the study. In this study, 60.4% of
the respondents stated UiTM was their first choice university while another 39.8%
of them did not state UiTM as their first choice university.
When we use too many items to measure a variable, we have to reduce them
into a manageable number before further analysis can be carried out. One of
the methods normally employed for data reduction is factor analysis. In the
factor analysis procedure, items which posses similar characteristic will be
grouped together under one component. So, instead of having to deal with 16
items, the researcher only deals with a few components resulted from the
procedure.
Example 10.3
Again we will use the data from Zainudin (2007) to illustrate the factor analysis
process. The data consists of 16 items to measure one construct namely the
corporate image. Here we want to reduce the number of items (if possible) by
grouping them together into a few components using factor analysis procedure.
The Principal Component Factor Analysis (PCA) with Varimax Rotation was
performed for the 16 items of the corporate image construct. The results (Table
10.9) indicate that the Bartlett’s Test of Sphericity was significant (Chi-square =
8498.069, p-value < 0.000). The measure of sampling adequacy by Kaiser-Meyer-
Olkin (KMO) is 0.932, and it has improved above 0.6.
Cont….
Prof Dr Zainudin Awang 229
UniSZA
Chapter 10: Data Analysis
[For information: the KMO value close to 1.0 and the significance value of
Bartlett’s Test close to 0.0 indicate the data at hand is adequate to proceed into
Factor Analysis].
Table 10.9: KMO and Bartlett's Test for Corporate Image Items
Kaiser-Meyer-Olkin Measure of
Sampling Adequacy. 0.932
Significance 0.000
This KMO value of 0.932 (Table 10.9) is excellent since it exceeds the
recommended value of 0.6 (Kaiser, 1974). The two measures (KMO value close to
1.0 and the Bartlett’s test significance value close to 0.0) suggest that the data is
appropriate to proceed with its reduction procedure. The next step for the
researcher is running Factor Analysis procedure in SPSS. Table 10.10 is the
output of a Factor Analysis for the 16 corporate image items.
cont…
The output in Table 10.10 shows that Factor Analysis procedure has extracted
three distinct dimensions with eigenvalues exceeding 1.0. In fact the researcher
could use up to four components since the eigenvalue for the fourth component
(0.93) is also close to 1.0. Remember, the more component extracted is better as
long as their eigenvalues are reasonably close to 1.0.
According to Igbaria et al. (1995), the item which loads greater than 0.6
under one component (say component 1) and loads lower than 0.35 under the
other components should be grouped into component 1. However, for the item
which has a loading of higher than 0.6 in one component and also loading higher
than 0.35 in other components is considered to have a positive cross-loading and
should be dropped from further analysis. However, the item which has a factor
loading lower than 0.6 under all components should also be dropped from the
analysis.
Example 10.4
Table 10.11 shows the factor loading for all 16 items under three components
namely component 1, component 2, and component 3. The study will select the
item using the above criteria to be grouped under the respective component. The
factor loading in bold indicates under which component the respective item falls.
In this case, item 1, 2, 3, 15, and 16 fall under component 1; item 4, 7, 8, 9, 10, 11,
and 12 falls under component 2; and item 5, 6, and 13 fall under component 3.
Meanwhile, item number 14 is excluded from the analysis since its loading is lower
than 0.6 for all components.
Once we have identified the items and their respective component, we can
transform the measurement of these items into one mean value representing their
respective component. In this case, component 1 is the mean of item 1, 2, 3, 15,
and 16; component 2 is the mean of item 4, 7, 8, 9, 10, 11, and 12; and
component 3 is the mean of item 5, 6, and 13.
cont…
Table 10.11: The Rotated Component Matrix for Corporate Image Items
Corporate Image Items Corporate Image Component
Component 1 Component 2 Component 3
7 University has good cooperation with its alumni 0.702 0.278 0.284
8 University has good cooperation with other universities 0.685 0.275 0.349
In fact, to make it clearly observable the item and its respective component, we
can instruct SPSS not to print the factor loading lower than 0.6 for all items. In this
case, the computer would printout only the factor loading higher than 0.6 for all
components as shown in Table 10.11a.
cont…
Table 10.11a: The Rotated Component Matrix for Corporate Image Items
The researcher could use suitable term to rename each component to reflect the
common items under that particular component. In this example, the items which
fall under component 1 are mostly related to the reputation of the university itself,
hence we could rename component 1 as “University Image”. The items which fall
under component 2 are mostly related to the academic excellence of the
university, hence we could rename component 2 as “Academic Image”. Finally, the
items which falls under component 3 are mostly related to the perception of
external organizations towards the university, hence the most appropriate name
for this component would be “External Recognition”.
Cont…
Prof Dr Zainudin Awang 234
UniSZA
Chapter 10: Data Analysis
However, if we do not want to rename the components for any reason, we can
maintain their variable names as component 1, component 2, and component 3 in
further analysis. Remember, instead of dealing with 16 items, the data has been
reduced to only three components for further analysis.
The next step is to determine the reliability measure for the measuring items under
each component. The popular measure for reliability is the Cronbach’s alpha. The
Cronbach’s alpha of 0.6 or higher for a component reflects the measuring items
under that particular component provide a reliable measure of internal
consistency.
Example 10.5
We compute the Cronbach’s alpha for each component in the above example. As
shown in Table 10.12, the values of Cronbach’s alpha are 0.909, 0.857, and 0.688
respectively for component 1, component 2, and component 3. It is important to
note here that all reliability measures have exceeded the minimum value of 0.6 as
recommended by Nunally (1978).
As shown in Table 10.12, there are seven items under component 1, five items
under component 2, and three items under component 3. The next step is to
transform the individual score of these seven items into its individual mean
representing component 1 for further analysis. The same procedure goes for other
components namely component 2 and 3 respectively. In the end, we only have
three values measuring the corporate image construct instead of 16 items.
After obtaining the mean scores for each component, the researcher needs to
show that the distribution of these data also do not depart from normality. As
stated earlier, one of the measures which reflect the distribution of data is
skewness. The measure of skewness between -1.0 to 1.0 indicates the data does
not depart from normality. Hence, the parametric statistical analysis can be
employed.
[Please note that another method to confirm the normality of data is the Kolmogorov-
Smirnov test statistics whereby Ho: data is normally distributed vs HA: data is not normally
distributed. We are looking for p-value higher than 0.05 so that the procedure would fail to
reject the null hypothesis to confirm that the data is normally distributed.]
Component 2
5.80 10.00 -0.19 -2.33 -0.64 -3.83
(Academic Image)
Component 3
5.00 10.00 -0.57 -6.86 0.57 3.41
(External Recognition)
Since all measures for the skewness are closer to 0.0 and within the range
between -1.0 to 1.0 as shown in Table 10.13, the study concludes that the
distribution of data is almost symmetry or bell-shaped. The bell-shaped distribution
indicates the data is normally distributed. Hence, the data obtained in the study
meets the required assumption for employing the parametric statistical analysis
that data comes from a normal distribution. Among the parametric statistical
analyses that can be employed are t-test, Z-test, and F-test depending on the
hypothesis test proposed for the study.
Before proceeding further with the analysis, the researcher needs to go back to his
research objectives in order to achieve the stated objectives, to address the stated
research questions, and also to test the stated hypotheses.
Example 10.6
Again we illustrate the analysis by using the data by Zainudin (2007). The
researcher has the following objective of the study, its related research
question, and the corresponding hypotheses.
Cont…
Testing the Hypothesis for Mean Differences between two Groups: The
independent two-sample t-test
Total 872
95% Confidence
Significance Mean Std. Error
t df Interval of the
Value (p) Difference Difference
Difference
Lower Upper
Since the P-value is higher than 0.05, the study fails to reject the null
hypothesis. Hence, the above research hypothesis is not supported.
Testing Hypothesis for Mean Differences for More than two Groups: The
one-way Analysis of Variance (ANOVA)
The ANOVA table (F-test) for equality of means among 5 income levels
Mean Significance
Parents’ Income Sum of Squares df Square F Value (p)
Between Groups 9.02 4 2.26 2.04 0.08
Since the P-value is higher than 0.05, the study fails to reject the null
hypothesis. Hence, the above research hypothesis is not supported.
Since the P-value is higher than 0.05, the study fails to reject the null
hypothesis. Hence, the above research hypothesis is not supported.
The ANOVA Table (F-Test) For Equality Of Means Among Five Education
Levels
Since the P-value is higher than 0.05, the study fails to reject the null
hypothesis. Hence, the above research hypothesis is not supported.
The ANOVA table (F-test) for equality of means among three duration
levels
Duration in the Sum of Mean Significance
program Squares df Square F Value (p)
Between 4.94 2 2.47 3.38 0.03
Groups
Within Groups 634.46 869 0.73
Since the P-value is lower than 0.05, the study manages to reject the
null hypothesis. Hence the above research hypothesis is supported.
Once the research hypothesis is supported, the study proceeds to do the
post-hoc analysis using the Least Significance Difference method (LSD).
The multiple comparison procedure is employed to determine which pair
has a significant difference in the mean score.
Mean
(I) Duration in the (J) Duration in the Difference Std. Significance
program program (I-J) Error Value (p)
between 4 to 5
0.10 0.07 0.16
between 2 to 3 semesters
semesters more than 5
-0.18 0.07 0.01
semesters
between 2 to 3
-0.10 0.07 0.16
between 4 to 5 semesters
semesters more than 5
0.08 0.07 0.25
semesters
between 2 to 3
0.18 0.07 0.01
more than 5 semesters
semesters between 4 to 5
-0.08 0.07 0.25
semesters
Prof Dr Zainudin Awang 241
UniSZA
Chapter 10: Data Analysis
Sometime the researcher develops certain theoretical framework for his study.
The theoretical framework, represented by a schematic diagram, indicates how he
theorizes the existence of relationship among the constructs involved in the study.
In the end he needs to determine whether the proposed relationship is significant
or not based on the data gathered.
Example 10.7
Zainudin (2007) also has the following theoretical framework to prove in his
study. The framework is developed based on what the researcher theorized
regarding the relationship among the constructs in his study.
Service
Quality
Students’
Satisfaction Students’
Loyalty
Corporate
Image
The study theorized that the two constructs namely service quality provided in a
university and the corporate image of a university have significant influence on
students’ satisfaction with the university. The students’ satisfaction, in turn, has a
significant influence on their loyalty towards the university in the future. In other
words, these two constructs (service quality and corporate image) have indirect
influence on customer loyalty through customer satisfaction.
In this study, there are two independent constructs namely service quality and
corporate image. The study also has an intervening (mediating) construct namely
students’ satisfaction, and the intended dependent construct namely students’
loyalty. The specific objectives for this study are as follows:
The related research questions for the study (an interrogative statement)
1 Does the service quality provided by the university influence students’
perceived satisfaction with their learning experience?
3 Does the service quality provided by the university influence students’ loyalty
intention towards their university?
Service
Quality
HA3
HA1
Students’
Satisfaction Students’
Loyalty
HA2
Corporate
Image HA4 HA5
Figure 10.2: Schematic diagram showing the theoretical framework and the
hypothesized relationship among the constructs in the study
Cont…
The statistical analysis for this type of relationship in social sciences is quite
complicated since the constructs involved are only hypothetical. These
constructs are latent (unobserved variables) and they are measured indirectly
using certain number of items. For example, the corporate image construct
alone is measured using 16 interval-scaled items as shown in Table 10.11.
Using the Factor Analysis procedure, those 16 items are grouped together into
three components. The same goes for other constructs (Table 10.14).
Table 10.14: The number of measuring items and its component for each construct
Number of Number of components
Name of construct measuring items after Factor Analysis
Corporate Image 16 3
Service Quality 34 3
Students’ Satisfaction 14 2
Students’ Loyalty 9 2
For information, one does not have to carry out the data reduction procedure
(Factor Analysis) if the number of items measuring the construct is less than
five. This is because the number of items is small and manageable.
The most popular method of analysis for this type of relationship is Structural
Equation Modelling (SEM), and the most convenient software for SEM is
AMOS Graphic. (AMOS stands for Analysis of Moments Structure). In AMOS
Graphic, the researcher draws the latent (unobserved) constructs followed by
the manifesting (observed) components.
Refer to Figure 10.2: The main constructs in this study as shown in the
theoretical framework are:
1 1 1
err9 err10
SERVICE
QUALITY 1 1
RetenN RecomM
err7
1
1
Process 1
STUDENTS' STUDENTS'
SATISFACTION LOYALTY
Outcome
1 1
1
Error2 Error1
err8
CORPORATE
IMAGE
1 1 1
Listed below is the definition of terms for all constructs in the proposed model
(Figure 10.3). Among the latent constructs in the model are:
2. Error1 - Estimated error to emerge when service quality, corporate image, and
students’ satisfaction are included in the equation to predict students’ loyalty
towards the university. This is called an error in the equation.
3. err9 - Estimated error to emerge when the intention to remain loyal and
increase transaction is employed to measure students’ loyalty towards the
university. This is called an error in the measurement.
6. Error2 - Estimated error to emerge when service quality and corporate image
are included in the equation to predict students’ satisfaction with the university.
This is called an error in the equation.
8. err8 - Estimated error to emerge when students’ satisfaction with the service
outcome is employed to measure their overall satisfaction with the university.
This is called an error in the measurement.
10. err1 - Estimated error to emerge when human elements in service delivery is
employed to measure service quality performance. This is called an error in the
measurement.
11. err2 - Estimated error to emerge when academic facilities at the university are
employed to measure service quality performance. This is called an error in the
measurement.
12. err3 - Estimated error to emerge when the infrastructure facilities is employed
to measure service quality performance. This is called an error in the
measurement.
13. CORPORATE IMAGE - This is another predictor construct in the model. The
study measured this construct using three manifesting variables namely image
of university (UNIVIMGE), image of academic activities (ACADMGE), and
external recognitions (RECOGN).
14. err4 - Estimated error to emerge when reputation of the university is employed
to measure corporate image. This is called an error in the measurement.
Once everything is in place, the researcher would run the AMOS model in
order to obtain the required statistical measures required in the study.
Automatically AMOS would give two distinct outputs namely the Standardized
Model and the Unstandardized Model. The Standardized Model is shown in
Figure 10.4 and the Unstandardized Model is shown in Figure 10.5.
err9 err10
SERVICE
QUALITY
.50 .70
RetenN RecomM
.05
err7 .46
.71 .84
.62
.40 .65
Process .79
STUDENTS' .66 STUDENTS'
.78 .83 LOYALTY
.91 SATISFACTION
Outcome
.21
Error2 .17 Error1
err8
CORPORATE
IMAGE
.82 .42
.90
Figure 10.4: The Path Diagram presents the strength and magnitude of
association among the constructs in the Study.
Table 10.16: The measures of strength and magnitude of association among the
constructs
Estimate of
Variables PATH Variables
Correlation
STUDENTS'_SATISFACTION ← CORPORATE_IMAGE 0.21
The path diagram also presents the measure of association between the two
independent constructs in the study namely service quality and corporate image.
The positive association between these two exogenous constructs indicates that
the increase in either construct (service quality provided by the university) would
likely increase the other construct (corporate image of a university) and vice-versa.
In other words, it’s just like killing two birds with one stone.
The Unstandardized Model in Figure 10.5 also presents the path diagram.
However, unlike the path diagram of the Standardized Model, the path diagram in
the Unstandardized Model presents the regression coefficients linking the
independent constructs to the dependent construct in the study. Refer to Figure
10.5 for the fitted model and table 10.18 for the corresponding text output.
In Figure 10.5, the covariance between Service Quality and Corporate image
is 0.25, the actual regression coefficient between service Quality and Customer
Satisfaction is 0.44, the error variance in measurement of “Humans” is 0.43. The
significance level (P-value) for each relation is given in the text output, as shown in
Table 10.18. These values can be used to decide the relevant hypothesis.
1 1 1
Process 1.00
STUDENTS' .67 ST UDENT S'
.25 LOYALT Y
1.25 SATISFACTION
Outcome
1 1
1 .44 .27
.23
.51
Error2 .42 Error1
err8
.13
CORPORATE
IMAGE
2.83 1.00
2.53
1 1 1
.49 .20 .60
Figure 10.5: The Path Diagram Presents the Regression Coefficients among
the Constructs
The path analysis among all constructs and variables in the model is
illustrated in Table 10.18. The symbol *** is given by AMOS to indicate the effect is
highly significant (P-value <0.001).
Table 10.18: The Path Analysis for All Constructs and Variables in the Model
Service
Quality
HA3
HA1
Students’
Satisfaction Students’
Loyalty
HA2
Finally, the study would like to test the main hypotheses in this research
(Figure 10.6). The following procedure would test the hypothesized links among
the main constructs in the study namely Service Quality, Corporate Image,
Students’ Satisfaction, and Students’ Loyalty as already stated above. The arrows
indicate that the study is interested to examine the significance of direct influence
of one construct to another construct.
HA1: Service quality has a significant and direct influence on customer satisfaction
Since the P-value is lower than 0.05, the above research hypothesis is supported.
The study concludes that service quality at the university has a significant and direct
influence on students’ satisfaction towards the university.
Since the P-value is lower than 0.05, the above research hypothesis is supported.
The study concludes that corporate image of the university has a significant and direct
influence on students’ satisfaction with the university.
HA3: Service quality has a significant and direct influence on customer loyalty
STUDENTS'_LOYALTY ← SERVICE_QUALITY
0.05 0.06 0.88 0.38
Since the P-value is higher than 0.05, the above research hypothesis is not
supported. The study concludes that service quality at the university does not have a
significant and direct influence on students’ loyalty towards the university.
HA4: Corporate image has a significant and direct influence on customer loyalty
STUDENTS'_LOYALTY ← CORPORATE_IMAGE
0.42 0.14 2.96 0.001
Since the P-value is lower than 0.05, the above research hypothesis is supported.
The study concludes that corporate image of the university has a significant and direct
influence on students’ loyalty towards the university.
HA5: Customer satisfaction has a significant and direct influence on customer loyalty
STUDENTS'_LOYALTY ← STUDENTS'_SATISFACTION
0.67 0.05 12.60 0.001
Since the P-value is lower than 0.05, the above research hypothesis is supported.
The study concludes that students’ satisfaction with the university has a significant and
direct influence on students’ loyalty towards the university.
Table 10.19: Summary of the Results of the Main Hypotheses in the Study
Result
No The Main Hypothesis Statement in the Study
Now the researcher needs to proceed into another chapter namely the
analysis of results, conclusion, and recommendation. In this chapter the
researcher needs to address every item of his original objectives, the research
questions, and the corresponding hypotheses.
Remember, at the end of any research the researcher needs to revise his
earlier objectives to ensure that they are achieved by the study. The research is
not meaningful if it fails to achieve the stated objectives.
The same thing also applies to the research questions; the researcher needs
to revise his research questions to ensure that their answers are properly
addressed based on the result of the study. The research is not meaningful if it
fails to address or to clarify the ambiguities surrounding the problem.
Exercises
b) What types of scalar data (nominal, ordinal, interval, or ratio) are typically
utilized in cross-tabulation analysis?
Policy 1: No advertising
Policy 2: Advertise in neighbourhood with circulars
Policy 3: Use circulars and advertise in local newspapers
Eighteen stores are randomly selected and divided at random into three groups
of six stores. Each group uses one of the three policies. Following the
implementation of the policies, sales figures are obtained for each of the stores
during a one-month period. The results are analyzed and obtained as follows.
Based on the SPSS output, answer the following questions.
a) Perform the normality test on sales figures based on the three advertising
policies. What can you conclude from the test? Is it appropriate to perform a
one-way analysis of variance? (Use α = 0.05).
c) From the post hoc multiple comparisons, indicate which pairs of advertising
policies are significantly different in terms of mean sales figures? State the
p-values.
Tests of Normality
Kolmogorov-Smirnov(a) Shapiro-Wilk
Statistic df Sig. Statistic Df Sig.
Sales 0.183 18 0.112 0.914 18 0.101
ANOVA
Sales
Sum of Mean
Squares df Square F Sig.
Between Groups 139.000 2
69.500
Within Groups 87.000 15 11.983 .001
5.800
Total 226.000 17
Scheffe’
Z value -2.105(a)
Asymptotic Sig. (2-tailed) P-value .035
location
FRONT MIDDLE REAR
Mean 6.067 2.067 3.733
Median 5.800 1.900 3.400
6) What types of scalar data (nominal, ordinal, interval, or ratio) are typically
utilized in cross-tabulation analysis?
7) Based on the frequency tables and summary statistics below; describe the
profile of the respondents.
Gender
Valid Cumulative
Gender Frequency Percent Percent Percent
Male 54 54.0 54.0
54.0
Female 46 46.0 46.0
100.0
Total 100 100.0 100.0
Stay
Valid Cumulative
Frequency Percent Percent Percent
On Campus 26 26.0 26.0
26.0
Off campus with friends 51 51.0 51.0
77.0
Off campus with family 23 23.0 23.0
100.0
Total 100 100.0 100.0
Descriptive Statistics
N Minimum Maximum Mean Std. Deviation
Age 100 13 27 20.72 2.656
Valid N (listwise) 100
Smoke
Valid Percent Cumulative
Frequency Percent Percent
Yes 19 19.0 19.0 19.0
No 60 60.0 60.0 79.0
Occasionally 21 21.0 21.0 100.0
Total 100 100.0 100.0
8) Based on the frequency table and the χ2 test of association results below,
write an analysis and conclusion.
a) Identify the target population and sample. Briefly explain how the
sampling frame is acquired.
c) Describe a suitable research design for the study and its purpose.
Chi-Square Tests
Asymp. Sig.
Value df (2-sided)
Pearson Chi-Square 9.813ª
2 .007
Likelihood Ratio 10.625
2 .005
Linear-by-Linear Association 2.574
1 .109
N of Valid Cases 100
a. 0 cells (.0%) have expected count less than 5. The minimum expected count is 8.74.
SEM begins with a theory whereby the researcher intends to test the
theorized relationships among research constructs in the study. This theoretical
relationship is modeled into a theoretical framework represented by a schematic
diagram. The schematic diagram presents the hypotheses of interest in the
research. The constructs involved are measured using a set of items in a
questionnaire. The measurement scale for each item should be either interval or
ratio. Ideally, the researcher should consider a minimum of four items to measure
each construct.
In research, latent constructs are the variables which the researcher could not
measure them directly since it is only a hypothetical concept of something. Thus,
the researcher cannot model these latent constructs in the Ordinary Least Square
regression models (OLS). Some examples of latent constructs are:
Service Quality, Customer Satisfaction, Job Satisfaction, Corporate Image,
Product Image, Customer Loyalty, Purchase Intention, Consumer Behavior,
Employee Soft Skills, Perceived usefulness, Relational bond, Financial bond,
Structural bond, Relationship quality, Attitudinal loyalty, and Behavioral loyalty.
e1 e2 e3 e4 e5 e6 e7 e8 e9 e10
1 1 1 1 1 1 1 1 1 1
1 1
1
X1 Y e11
Note: X1 and Y are latent constructs. In Amos graphic, latent constructs are
represented by ellipses. The latent construct X1 is measured using five items
namely X11 to X15, while latent construct Y is also measured using five items
namely Y1 to Y5. These measured or observed variables are represented by
rectangles in AMOS model.
e1 e2 e3 e4 e5 e6 e7 e8 e9 e10
1 1 1 1 1 1 1 1 1 1
X1 Y
When the relationships among variables are complex and the variables are
unobserved or latent constructs, modeling the inter-relationship among constructs
are almost impossible using OLS. However, one could employ SEM to model and
estimate the strength of relationship between constructs.
AMOS can model and analyze the multiple regression equations for latent
constructs simultaneously. Figure 11.3 shows how three simple regression
equations and one multiple regression equation is modeled simultaneously. At the
same time the covariance between the exogenous constructs are also estimated.
e1 e2 e3 e4 e5
1 1 1 1 1
X1
1 1
e6 X21 e21 Y1 e16
1
1 1 1
e7 X22 Y2 e17
1 1
e8 X23 X2 Y Y3 e18
1 1
e9 X24 1 Y4 e19
1 1
e10 X25 Y5 e20
X3
1
Figure 11.31: The Structural Equation Model for the Latent Constructs (Y On
X1, X2, and X3)
AMOS could analyze the mediating effects of latent construct in a model. The
theoretical model is illustrated in Figure 11.4. In the model, exogenous construct
X1 has five response items; mediating construct X2 also has five response items,
while endogenous construct Y has three response items respectively. Here, X2 is
modeled as a mediator construct. The researcher is interested to determine
whether X2 mediates the relationship between X1 and Y.
e6 e7 e8 e9 e10
1 1 1 1 1
1
X2 e14
1
1 Y1 e11
1
X1 Y Y2 e12
1 1
1 Y3 e13
e1 e2 e3 e4 e5
Figure 11.42: Modeling the Mediating Effect of Latent Construct X2. The
Illustration of the Exogenous Variable (X1), Mediating Variable (X2),
And Endogenous Variable (Y)
AMOS could model and analyze the effects of moderator variable in the
model. The theoretical model is illustrated in Figure 11.5. In the model, exogenous
construct X1 has five response items and endogenous construct Y also has five
response items. In the model, M is the moderator variable. The researcher is
interested to determine whether M moderates the relationship between X1 and
Y.
e1 e2 e3 e4 e5 e6 e7 e8 e9 e10
1 1 1 1 1 1 1 1 1 1
1
X1 Y e11
1. Unidimensionality
i) For a newly developed scales, the factor loading for an item should be
0.5 or higher
ii) For already established scales, the factor loading for an item should
be 0.6 or higher
The item-deletion process should be made one item at a time with the
lowest item to be deleted first. After an item is deleted, the researcher
needs to re-specify and estimate the new measurement model. The
process continues until the unidimensionality requirement is achieved.
2. Validity
3. Reliability
In SEM, there are a series of goodness-of-fit indexes that reflect the fitness of the
model to the data at hand. At the moment, there is no agreement among the
researchers which fitness indexes should be reported. Hair et al. (1995) and
Holmes-Smith (2006) recommend the use of at least three fit indexes by including
at least one index from each category of model fit. The three fitness categories are
absolute fit, incremental fit, and parsimonious fit.
The researcher could choose at least one fitness index from each category to
report depending on which literature is being referred. The information concerning
the fitness index category, their level of acceptance, and comments is presented in
Table 11.3.
Table 11.3: Index Category and the Level of Acceptance for Every Index
Absolute fit RMSEA RMSEA < 0.08 Range 0.05 to 1.00 acceptable
Absolute fit GFI GFI > 0.90 GFI = 0.95 is a good fit
2. Incremental fit AGFI AGFI > 0.90 AGFI = 0.95 is a good fit
Incremental fit CFI CFI > 0.90 CFI = 0.95 is a good fit
3. Parsimonious fit Chisq/df ChiSq/df < 5.0 The value should be less than 5.0.
***The indexes in bold is recommended since they are highly reported in literatures.
The types of fitness indexes employed and the acceptable cut-off values reported
by researchers may vary depending on which literatures support they are referring.
However, the following table presents the literature support for the selected fitness
indexes.
Table 11.4: The Literature Support for the Respective Fitness Index
*** One could ignore the absolute fit index of minimum discrepancy chi-square if the
sample size obtained for the study is greater than 200 (Hair et al., 1996; Joreskog and
Sorbom, 1996).
e1 e2 e3 e4 e5 e6 e7 e8 e9
1 1 1 1 1 1 1 1 1
Wl 1 Wl 2 Wl 3 Wl 4 Wl 5 Wl 6 Wl 7 Wl 8 Wl 9
Work
Load
Reference Point
The CFA procedure, the researcher should instruct AMOS to calculate the
standardized estimate of factor loading and squared multiple correlation for every
item by clicking the respective box in the analysis properties menu. The
standardized estimates indicate the factor loading for each item in a measurement
model.
The AMOS output showing the factor loading and squared multiple
correlation for every item in a measurement model are presented in Figure 11.7.
e1 e2 e3 e4 e5 e6 e7 e8 e9
Wl 1 Wl 2 Wl 3 Wl 4 Wl 5 Wl 6 Wl 7 Wl 8 Wl 9
.71 .77
.73 .80 .82 .83 .65
.70 .75
Response
Item R2
Work
Factor Loading Load Latent construct
Figure 11.7: The Standardized Estimate of Factor Loading for Every Item in
The Measurement Model for Employee’s Workload
Figure 11.7 presents the factor loading for each item in a measurement
model to measure the workload construct. The factor loading for a particular item
is shown near the arrow pointing to the respective item, while the value shown
above each response item is the squared multiple correlation or R2 for that
particular item. The literature suggests that any measuring item having a factor
loading less than 0.6 and an R2 less than 0.4 should be deleted from the
measurement model. However, it depends on the decision by the researcher if he
has strong reason not to drop any of the items in the model.
To solve the correlated errors or redundant items, the researcher could choose
one of the following two choices:
Choice 1:
i) Delete one item at a time (choose the lower factor loading to delete first)
ii) Re-specify the model and run (repeat step 1 to 7)
Choice 2:
8) Obtain the Cronbach’s Alpha, CR, and AVE for the measurement model
Note: The acceptable value of factor loading, the level of fitness index, and also the
method of modification to the measurement model varies among the literatures.
Let say the schematic diagram of the model under study is given in Figure
11.8 and the measurement model is given in Figure 11.9.
X1
X2
Figure 11.9 presented the measurement model for all constructs involved in
the study based on the schematic diagram shown in Figure 11.8. The CFA
procedure is executed based on the measurement model in Figure 11.9. The
researcher instructs AMOS to calculate the factor loading and R2 for every item;
and also the strength of correlation between the constructs.
The item having factor loading less than 0.6 should be deleted from the
measurement model. If the correlation between two exogenous is higher than
0.85, then discriminant validity is not achieved. The output for CFA is presented in
Figure 11.10.
1
X11 e1
1
1
X12 e2
1
X1 X13 e3
1
X14 e4
1
X15 e5
1
X21 e6
1
1
X22 e7
X2
1
X23 e8
1
X24 e9
1
Y1 e10
1
1
Y2 e11
Y
1
Y3 e12
1
Y4 e13
Figure 11.9: The Measurement Model for All Latent Constructs in The Study
The CFA results showing the factor loading and the squared multiple correlations
(R2) for every item. At the same time, the correlation between constructs is
presented in Figure 10.
.38
Factor Loading
X11 e14
2
.62 .32 R
X12 e15
Correlation
.57 .82
.91
X1 X13 e16
.88 .78
X14 e17
.92 .85
X15 e18
.87
.61
X21 e19
.93
.91
.96
X22 e20
.56
X2 .66
.44
X23 e21
.53
.28
X24 e22
.47
.91
Y1 e10
.96
.95
.98
Y2 e11 Fitness Indexes
1. ChiSq = 277.471
Y .94
.89
2. df = 62
Y3 e12 3. ChiSq/df = 4.475
4. GFI = .879
.63 5. AGFI = .823
.80
6. CFI = .943
Y4 e13 7. RMSEA = .105
Figure 11.10: The Factor Loading For All Items of the Respective Construct
As shown in Figure 11.10, the factor loadings for item X12 (from construct X1),
and item X24 (from construct X2) are below 0.6. Thus, the researcher has to delete
these two items and the new measurement model has to be re-specified. These
two items were deleted and the re-specified model is shown in Figure 11.11.
.36
X11 e1
.60
.83
.91
X1 X13 e3
.88 .77
X14 e4
.93
.86
X15 e5
.86
.60
X21 e6
.93
.93
.97 X22 e7
.56
X2 .65
.42
X23 e8
.46
.91
Y1 e10
.96
.95
Figure 11.11: The Factor Loading For Other Items After X12 and X24 Were
Deleted
The factor loading for all items in Figure 11.11 are above 0.6. Thus no more
item deletion is required for this measurement model. Now, the unidimensionality
requirement has been achieved. The fitness indexes for the model are assessed in
Table 11.5.
Table 11.5: The Summary of Fitness Indexes for the Measurement Model
If the researcher is already satisfied with the fitness indexes obtained, then
the CFA for measurement model is done at this stage. The next step is modeling
the structural model, which will be discussed in next chapter. Otherwise, if the
researcher is still not satisfied with AGFI (less than 0.9), he could examine for
redundant items in the measurement model through the MI table produced by
AMOS. The MI table for the above measurement model is presented in Table
11.6.
Note: The MI value above 15 is considered high and that particular item is deemed to be
redundant.
Table 11.6: The Modification Indices Presents the Covariance between Each
Pair of Items
Parameter
M.I. Comment
Change
e12 <--> e13 26.789 0.127 MI indicates item e12 and e13 are correlated
e10 <--> e13 4.643 -0.044
e5 <--> e7 4.568 -0.032
e1 <--> e7 10.890 0.085
The MI = 26.789 (in bold) is considered high which indicate that item Y3 and
item Y4 are redundant (e12 = item Y3 and e13 = item Y4). As has been discussed
Prof Dr Zainudin Awang 286
UniSZA
Chapter 11: The Structural Equation Modeling
earlier, the researcher has two choices to remedy this problem. The two choices
are:
Choice 1: to delete one of these two redundant items and re-specify the
measurement model.
Choice 2: to set these two correlated errors to be “free parameter estimate” and
re-specify the measurement model.
Let say, the researcher elects to employ choice 2 for this particular case. The
result is presented in the following diagram (Figure 11.12).
.36
X11 e1
.60
.83
.91
X1 X13 e3
.88 .77
.93 X14 e4
.86
X15 e5
.86
.60
X21 e6
.93
.93
.97
X22 e7
.56
X2 .65
.42
X23 e8
.46
.91
Y1 e10
.96
.96
Figure 11.12: The Re-Specified Model after a Pair of E12 and E13 is Set as
“Free Estimate”
The fitness indexes for the model are assessed in the following table (Table 11.7).
Name on
Name of index category Index value Comments
index
1. Absolute fit RMSEA 0.065 Achieved the requirement
Absolute fit GFI 0.949 Achieved the requirement
2. Incremental fit CFI 0.985 Achieved the requirement
3. Parsimonious fit Chisq/df 2.344 Achieved the requirement
Note: Look how the fitness indexes have improved after the model re-specification was made.
At this stage, the CFA for the measurement model is completed. And as
usual, the business now is to address the issues of unidimensionality, validity, and
reliability before moving to next stage (modeling the structural model).
Table 11.8: The Suggested CFA Results Reporting For The Measurement
Model
CFA Results Reporting
Cronbach CR AVE
Factor
Construct Item Alpha (Above (Above
Loading
(Above 0.7) 0.5) 0.6)
X1 X11 0.60 0.893 0.904 0.707
X13 0.91
X14 0.88
X15 0.93
X2 X21 0.93 0.869 0.894 0.743
X22 0.97
X23 0.65
Y Y1 0.96 0.946 0.955 0.844
Y2 0.98
Y3 0.94
Y4 0.78
Discriminant Validity
Construct X1 X2 Y
X1 0.83
X2 0.60 0.85
Y 0.56 0.46 0.92
The diagonal values in bold is the square root of AVE while other values are
the correlation between the respective constructs. The discriminant validity is
achieved when a diagonal value in bold is higher than the values in its row and
column.
Once the issues of unidimensionality, validity, and reliability have been addressed,
the researcher needs to examine the normality assessment for the data at hand
before proceed to modeling the structural model. Using the final measurement
model, select the output box named “test for normality and outliers” in order to
assess the distribution for every variable in a dataset. The following table presents
the resulted output from the abovementioned procedure. Table 11.10 presents the
normality assessment for every item involved in the measurement model.
output tabulates the list of outliers once the box “test for normality and
outliers” is selected.
2. The value of multivariate kurtosis reflects the multivariate normality
distribution for data set. The value of multivariate kurtosis should be lower
than 50.0; otherwise the assumption of multivariate normality is not satisfied.
In the case where the assumption is not met, the researcher should examine
the data outliers toward the model. The deletion of few extreme outliers might
help improve the multivariate normality. The outlines in AMOS are shown by
the Mahalanobis Distance. AMOS output will tabulate the list of observations
and their respective distance from the centroid. The centroid is the centre of
data distribution.
Otherwise, if the deletion process does not work the researcher needs to
employ bootstrapping technique in order to confirm multivariate normality. If
the bootstrapping method still fails to confirm multivariate normality of data
set, the researcher should abandon using SEM.
First of all, refer to the theorized inter-relationships between the constructs. Most
of the times, the inter-relationships between constructs is presented in the
schematic diagram as shown below. Based on the schematic diagram, draw the
AMOS graphic for the structural model as shown in Figure 11.12.
X1
X2
1
e1 X11
1
1
e3 X13 X1 e14
1
1
Y1 e10
e4 X14 1 1
1
1
Y2 e11
e5 X15
Y
1
1 Y3 e12
e6 X21
1
1
1 Y4 e13
e7 X22
X2
1
e8 X23
Based on the structural model in Figure 11.12, the researcher could instruct AMOS
to compute the values for the Standardized Estimate and also the Unstandardized
Estimate for the model. The values obtained under Standardized Estimate are
presented in Figure 11.13, while the values under the Understandardized Estimate
are presented in Figure 11.14.
.36
e1 X11
.60
.83
.91
e3 X13 X1 e14
.88 .91
.77 Y1 e10
e4 X14 .45 .96
.93
.98 .96
.86 Y2 e11
e5 X15
.60 .34 Y
.88
.86
.19 .94 Y3 e12
e6 X21
.93 .33
.78 .61
.93 Fitness Indexes Y4 e13
e7 X22 .97
1 ChiSq = 93.747
2 df = 40
.42 X2 3 ChiSq/df = 2.344
e8 X23 4 GFI = .949
.65
5 AGFI = .917
6 CFI = .985
7 RMSEA = .065
Note:
The standardized beta estimate of X1 is 0.45, and the standardized beta estimate
for X2 is 0.19
The measure of correlation between construct X1 and X2 is 0.60.
.87
1
e1 X11
1.00
.23 .76
1 1.51
e3 X13 X1 .49
e14 .11
.28 1.40 1
1
Y1 e10
e4 X14 .68 1 1.00
.05
1.59
.20 1.03 1
1
Y2 e11
e5 X15
.45 Y .19
.20 1
1 .19 1.09 Y3 e12
e6 X21
1.00 .93 .14
1.12
.09 1
1 Fitness Indexes Y4 e13
e7 X22 1.00
1 ChiSq = 93.747
2 df = 40
1.07 X2 1.19 3 ChiSq/df = 2.344
1
e8 X23 4 GFI = .949
.81 5 AGFI = .917
6 CFI = .985
7 RMSEA = .065
Table 11.11: The Fitness Index Assessment for the Structural Model in
Figure 11.14
AMOS would give two sets of text output namely the standardized estimate, and
unstandardized estimate. The text output explaining results in Figure 11.13 are
presented in Table 11.11, Table 11.12, and Table 11.13. While the text output
explaining the results in Figure 11.14 are presented in Table 11.14, Table 11.15,
and Table 11.16.
Y <--- X2 0.19
Y <--- X1 0.45
Table 11.13: The Correlation Estimate for Each Pair of Exogenous Construct
Variable 2
Estimate (R )
X11 0.381
X24 0.280
X23 0.440
X22 0.914
X21 0.868
The definition for the squared multiple correlations are given below.
Y 0.341
It is estimated that the predictors of Y explain 34.1 percent of its variance. In other
words, the error variance of Y is approximately 65.9 percent of the variance of Y
itself.
Y4 0.633
It is estimated that the predictors of Y4 explain 63.3 percent of its variance. In other
words, the error variance of Y4 is approximately 36.7 percent of the variance of Y4
itself.
X15 0.852
It is estimated that the predictors of X15 explain 85.2 percent of its variance. In
other words, the error variance of X15 is approximately 14.8 percent of the variance
of X15 itself.
The actual
S.E. C.R. P-value
beta values
Y <--- X1 0.662 0.106 6.233 ***
The actual
S.E. C.R. P-value
beta values
Y <--- X2 0.188 0.062 3.052 0.002
Definition for beta 0.188: When X2 goes up by 1unit, Y goes up by 0.188 unit.
Definition for Critical Ratio for regression weight: Dividing the regression
weight estimate by the estimate of its standard error gives
z = 0.188/0.062 = 3.052. In other words, the regression weight estimate is 3.052
standard errors above zero.
The actual
S.E. C.R. P
beta values
Y <--- X1 0.662 0.106 6.233 ***
Definition for beta 0.662: When X1 goes up by 1 unit, Y goes up by 0.662 units.
Definition for Critical Ratio for regression weight: Dividing the regression
weight estimate by the estimate of its standard error gives z = 0.662/0.106 =
6.233. In other words, the regression weight estimate is 6.233 standard
errors above zero.
Table 11.16: The Regression Weight for each item. Note the reference point.
Dividing the covariance estimate by the estimate of its standard error gives
z = 0.480/0.065 = 7.357
In other words, the covariance estimate is 7.357 standard errors above zero.
The probability of getting a critical ratio as large as 7.357 in absolute value is less
than 0.001. In other words, the covariance between X1 and X2 is significantly
different from zero at the 0.001 level.
Table 11.18: The Variance Estimate for Each Variable in the Model
Variable Variance
S.E. C.R. P-value
Estimate
X2 1.200 0.112 10.695 ***
X1 0.516 0.087 5.942 ***
e14 0.749 0.067 11.133 ***
e6 0.183 0.030 6.081 ***
e7 0.110 0.027 4.072 ***
e8 1.039 0.086 12.028 ***
e9 1.352 0.110 12.307 ***
e1 0.839 0.069 12.075 ***
e2 1.212 0.099 12.188 ***
Variable Variance
S.E. C.R. P-value
Estimate
e3 0.241 0.028 8.678 ***
e4 0.268 0.028 9.572 ***
e5 0.210 0.027 7.691 ***
e10 0.108 0.012 8.830 ***
e11 0.059 0.010 5.785 ***
e12 0.171 0.017 9.823 ***
e13 0.870 0.072 12.034 ***
Table 11.18 presents the variance estimate for every variable in the model.
The probability or p-values for all variables are less than 0.001 which indicate that
the variances for all variables are significantly different from zero.
Mediating variable is the variable that mediates the effect from an independent
variable to its dependent variable. If the direct effect from variable X to variable Y
does not exist, instead the effect exist indirectly through another variable M; then in
this case M is a mediating variable. The position of mediating variable in the model is
illustrated below.
Mediating
Variable
Independent Dependent
Variable Variable
First of all, the direct effect of independent variable on dependent variable is significant.
However, when the mediating variable enters the model, the direct effect is reduced but
the indirect effect is significant. If direct effect is reduced but still significant, then the
“partial mediation” occurs. However, if the direct effect is no longer significant, then the
“complete mediation” is said to occur.
303
In this model, the researcher can assess the direct effect as well as the indirect
effect of X1 on Y. If the direct effect of X1 on Y is not significant, but instead, the indirect
effect through X2 is significant, then X2 is a mediating variable in the X1 to Y relationship.
0,
e1
X1 Y
0, 1.07
e1
1
4.90, 1.36 3.87
.36
X1 Y
Figure 12.2: The Result of Direct Effect Of X1 (The Beta Coefficient Is 0.36)
304
Table 12.1: The direct effect of X1 on Y is significant (Beta coefficient 0.361)
Secondly, we include the mediator X2 into the model as shown in Figure 12.3.
Now we assess the direct effect of X1 on Y as shown in Figure 12.4. The output in Table
1a shows B1 is reduced from 0.361 to 0.103, and the hypothesis is no longer significant.
Here, the requirement for complete mediation is met. However, we need to test the
hypothesis for B2 and B3 to confirm.
1
X2 e2
e1
1
X1 Y
305
3.10
0, 1.12
X2 1
e2
0, .98
.53 .31
e1
1
4.90, 1.36 3.01
.10
X1 Y
Refer to Table 12.2. The p-value for the test is 0.062. Thus, the regression weight
for X1 in the prediction of Y is not significant at the 0.05 level. In other words, the
direct effect of X1 on Y is not significant.
2. Hypothesis 2: X1 has significant and direct effects on X2
Refer to Table 12.2. The p-value is less than 0.001. Thus, the regression weight for
X1 in the prediction of X2 is significant at the 0.001 level. In other words, the direct
effect of X1 on X2 is significant.
306
3. Hypothesis 3: X2 has significant and direct effects on Y
Refer to Table 12.2. The p-value is less than 0.001. Thus, the regression weight
for X2 in the prediction of Y is significant at the 0.001 level. In other words, the
direct effect of X2 on Y is significant.
e14
1
Y1 e10
1 1
1
Y2 e11
Y
X1 1
Y3 e12
1 1
Y4 e13
X15 X14 X13 X11
1 1 1 1
e5 e4 e3 e1
307
.78
e14 .11
1
Y1 e10
1 1.00
.06
.47 1.03
1
Y2 e11
.87 Y .17
1.10
X1 1
Y3 e12
1.15
1.61 1.00 .87
1.42 1.53
1
Y4 e13
Fitness Indexes
X15 X14 X13 X11 1.ChiSq = 55.676
2.df = 19
1 1 1 1 3.ChiSq/df = 2.930
.19 .28 .23 .88 4.GFI = .956
5.AGFI = .917
e5 e4 e3 e1 6.CFI = .986
7.RMSEA = .078
The analysis for mediation begins by showing that the direct effect of X 1 on Y is
significant. The direct effect is measured through beta coefficient. In this case, B 1 is
significant (Table 12.3). When the mediating variable X2 enters the model, the value of
beta coefficient for X1 is expected to decrease, or in other words the direct effect of X 1
on Y would be reduced. The model when the mediator X2 is enters is presented in
Figure 12.7.
308
e6 e7 e8 e9
1 1 1 1
1
e15 X2
e14
1
Y1 e10
1 1
1
Y2 e11
Y
X1 1
Y3 e12
1
1
Y4 e13
X15 X14 X13 X11
1 1 1 1
e5 e4 e3 e1
In the above diagram, X1, X2, and Y are latent constructs. The study is interested
to prove that the construct X2 is mediating the relationship between construct X1 and Y.
The output for unstandardized estimates for the model is given in Figure 12.8,
while the text output is presented in Table 12.4.
309
.45
.19 .09 1.07 1.39
e6 e7 e8 e9
1 1 1 1
1.00 .81
1.00
.64
1 .75
.77 e15 X2
e14 .11
1
.94 .19 Y1 e10
1 1.00
.06
.48 1.03
1
Y2 e11
.68 Y .17
1.10
X1 1
Y3 e12
1.59 1.15
1.00 .87
1.40 1.51
1
Y4 e13
Fitness Indexes
X15 X14 X13 X11 1.ChiSq = 151.309
2.df = 50
1 1 1 1 3.ChiSq/df = 3.026
.19 .28 .23 .87 4.GFI = .929
5.AGFI = .889
e5 e4 e3 e1 6.CFI = .972
7.RMSEA = .080
Figure 12.5: The AMOS Output Showing Regression Coefficient between The
Constructs
310
Table 12.4: The Multiple Regression Weights
Refer to Table 12.4: The p-value for the test is 0.001. Thus, the regression weight
for X1 in the prediction of Y is significant at the 0.001. In other words, the direct
effect of X1 on Y is significant.
Refer to Table 12.4: The p-value for the test is 0.001. Thus, the regression weight
for X1 in the prediction of X2 is significant at the 0.001. In other words, the direct
effect of X1 on X2 is significant.
Refer to Table 12.4: The p-value for the test is 0.002. Thus, the regression weight
for X2 in the prediction of Y is significant at the 0.01. In other words, the direct
effect of X2 on Y is significant.
311
CHAPTER
ANALYZING THE MODERATING VARIABLES
13
Moderating variable is the variable that moderates the strength of causal effects from
an independent variable X to its dependent variable Y. Let M be the moderator
variable in the X-Y relationship. The moderation effects of M would “alter” the causal
effects of independent variable X on dependent variable Y.
We begin by assuming that the causal effect of independent variable X on its dependent
variable Y is significant. When the moderating variable M is introduced into the model,
the causal effects would change due to some interaction effect between independent
variable X and moderator variable M. The effects of X on Y could be no longer
significant or it could be more significant after interaction. In other words, the moderator
variable M is “moderating” the strength of X-Y relationship.
Example: Let X be the amount of environmental news concerning the safe and
clean environment. The news intends to make public aware of environmental
degradation and that they save the environment by switching to green products. Let Y
be the respondents’ intention to consume green products, and let M be their level of
education. If the effect of environmental news (X) in educating the public to purchase
green products (Y) is more visible among higher educated group compared to low
educated group, then we can say that level of education (M) is the moderating variable
in the relationship between independent variable and dependent variable.
Figure 13.1 illustrates the position of moderating variable (Education) in the in the
relationship between independent variable (Environmental Awareness Campaign) and
the resulting dependent variable (Intention to Purchase Green Products).
311
Environmental Intention to
Awareness Purchase
Campaign Green Products
Level of
Education
As has been said earlier, although moderation implies a weakening of a causal effect, a
moderator can also reverse or enhance that particular effect. A key part of moderation
is the measurement of causal effect X on Y for different values of moderating variable
M. In statistics, the effect of X on Y for a fixed value of M is referred as the simple effect
of independent variable on its dependent variable.
312
Types of moderation
1) Complete Moderation
2) Partial Moderation
3) No Moderation
X
e1
1
M Y
XM
The moderating effects of variable M is said to exist in the X-Y relationship if the XM - Y
relationship (B3) or Hypothesis 3 is significant while the M-Y relationship (B2) or
Hypothesis 2 is not significant.
314
ANALYZING THE MODERATING EFFECTS FOR OBSERVED VARIABLES
Suppose the researcher is interested to assess the moderating effects of age of workers
in the relationship between the monetary incentives given to them and their monthly
productivity. The AMOS model illustrating the researcher’s theoretical argument is given
in Figure 3. All variables in the model are directly observed.
Incentives
e1
1
Age Productivity
I.A
Figure 13.3: The Model Representation in AMOS Graphic for Testing Age as
Moderator Variable
315
.48
Incentives .60
.59 e1
.24 1
.49
.68 -.88
.20
I.A
Figure 13.4: The AMOS Output Showing the Beta Coefficients, Variance, and
Covariance
Table 13.1 show the p-value is less than 0.001 (***). Thus, the regression weight for
Incentives in the prediction of Productivity is significant.
In other words, the hypothesis that the causal effects of incentives on
productivity are significant has been supported.
316
Hypothesis 2: The Workers’ age level has significant effects on their productivity
Table 13.2: Testing the causal effects of Workers’ Age on their Productivity
Table13.2 show the p-value is 0.072. Thus, the regression weight for Age in the
prediction of Productivity is not significant.
In other words, the hypothesis that the causal effects of age on productivity are
significant is not supported.
Table 13.3 show the p-value is 0.001 (***). Thus, the regression weight for I.A in the
prediction of Productivity is significant.
In other words, the hypothesis that the moderating effects of workers’ age (M) on
the relationship between incentives (X) and their productivity (Y) are significant
has been supported.
Note: The type of moderation that occurs in this case is partial moderation since
the hypothesis for the main effect is also significant.
317
MODELING THE MODERATING EFFECTS FOR LATENT CONSTRUCTS
Analyzing the moderating effects for the model with latent constructs is very
complicated. The normal modeling procedure in Regression Analysis (using interaction
terms between independent and moderator variable) is not practical with latent
constructs since it would cause problems with model convergence as well as distortion
of standard errors. In the end, it resulted in model misfit and the procedure stops.
The position of moderator in the relationship between X1 and Y is shown in Figure 13.5.
e1 e2 e3 e4 e5 e6 e7 e8 e9 e10
1 1 1 1 1 1 1 1 1 1
1 1
1
X1 Y e11
Moderator
Figure 13.5: Modeling the Moderator Variable in the Path Between X1 and Y
318
ANALYZING THE MODERATOR FOR LATENT CONSTRUCTS: THE MULTI-GROUP
CFA
1) Split data equally into two groups based on the median of the moderator variable.
2) Save data into two separate files. Rename the two groups as group 1 and group 2.
3) Obtain the final model for the study. Identify the path of interest where the effect of
moderator variable is to be assessed.
4) Develop two separate AMOS models. Rename as model 1 and model 2 respectively.
7) Execute model 1 (constrained model) using both data 1 and data 2 separately.
8) Execute model 2 (unconstrained model) using both data 1 and data 2 separately.
9) Obtain the difference of Chi-Square value between the constrained model and the
unconstrained model separately for data 1 and data 2.
10) If the Chi-Square value between constrained and unconstrained model differs by
more than 3.84, then the moderator variable has a significant moderating effect on
the relationship between the constructs in the selected path.
Example: Suppose that we are modeling the effect of two independent variables
namely X1 and X2 on a dependent variable namely Y (Figure 13.6). One of the
objectives of this research is to assess the moderation effect of moderator variable
namely education in the relationship path between X1 and Y.
319
e1 e3 e4 e5
1 1 1 1
e14
X1 1
Y1 e10
1 1
1
Y2 e11
Y
1
Y3 e12
1
Y4 e13
X2
1 1 1 1
e9 e8 e7 e6
Figure 13.6: The Path Of Interest To The Researcher Here Is From X1 To Y Where
The Effect Of Moderator Variable Namely Education Is To Be
Examined.
The path of interest where the moderation tests to be carried out is shown in
Figure 13.6. First of all, sort data in ascending order based on respondents’ level of
education. Then the data is split equally based on the median of education and saved
into two separate data files. Data 1 is renamed as low education group, while data 2 is
renamed as high education group.
320
Secondly, put a constraint on the path of interest in the model to be equal to “1” as
shown in Figure 13.7. This model is renamed as the constrained model.
e1 e3 e4 e5
1 1 1 1
constrained e14
X1 1
Y1 e10
1 1
1
1
Y2 e11
Y
1
Y3 e12
1
Y4 e13
X2
1
1 1 1 1
e9 e8 e7 e6
Figure 13.7: The Constrained Model: The Coefficient In The Path Of Interest (X1 To
Y) Where The Effect Of Moderator Variable To Be Examined Is
Constrained To 1.
Thirdly, using the same model, remove the constraint in the path as shown in Figure
13.8. This model is renamed as the unconstrained model.
321
e1 e3 e4 e5
1 1 1 1
Parameter
X11 X13 X14 X15
Unconstrained
1
e14
X1 1
Y1 e10
1 1
1
Y2 e11
Y
1
Y3 e12
1
Y4 e13
X2
1 1 1 1
e9 e8 e7 e6
Figure 13.8: The Unconstrained Model: The Coefficient In The Path Of Interest (X1
To Y) Where The Effect Of Moderator Variable To Be Examined Is
Not Constrained.
Finally, obtain the estimate for both the constrained model and also the
unconstrained model using the same dataset (data 1: low education group). The output
is presented in Figure 13.9 for the constrained model and in Figure 13.10 for the
unconstrained model. The procedure for testing moderation is carried out as shown in
Table 13.4.
322
.87 .19 .24 .25
e1 e3 e4 e5
1 1 1 1
Constrained Slope
1.78
1.00 1.65
.73
1.76
e14 .09
X1 1
Y1 e10
1 1.00
1.00
.39 .04
1
1.02 Y2 e11
.43 Y 1.09
.20
1
Y3 e12
.07 1.05
1.31 .72 .12
1
Y4 e13
X2 Fitness Indexes
1. ChiSq = 107.979 Chi-Square
1.00 2. df = 50
.62 .83 1.01 and its df
3. ChiSq/df = 2.160
4. GFI = .905
5. AGFI = .851
X24 X23 X22 X21 6. CFI = .970
7. RMSEA = .086
1 1 1 1
1.41 .96 .04 .18
e9 e8 e7 e6
.51
Figure 13.9: Data 1: The Constrained Model. The Value of Chi-Square Is 107.979,
Df=50
323
.86 .19 .24 .25
e1 e3 e4 e5
1 1 1 1
Unconstrained Slope
X11 X13 X14 X15
1.55
1.00 1.44
.72
1.54
e14 .09
X1 1
Y1 e10
1 1.00
.69
.51 .04
1
1.03 Y2 e11
.49 Y 1.09
.20
1
Y3 e12
.14 1.06
1.31 .72 .12
1
Y4 e13
X2 Fitness Indexes
1. ChiSq = 104.520
.62
1.00 2. df = 49 Chi-Square
.83 1.01 3. ChiSq/df = 2.133 and its df
4. GFI = .906
5. AGFI = .851
X24 X23 X22 X21 6. CFI = .971
7. RMSEA = .085
1 1 1 1
1.41 .96 .05 .18
e9 e8 e7 e6
.51
Figure 13.10: Data 1: The Unconstrained Model. The Value of Chi-Square Is 104.520, Df=49
324
Table 13.4: The Measures Required For Performing Moderation Test Using Data 1
Not Supported
Ha: Respondent’s education moderate the causal effects of X1 on Y
The procedure for performing the test of moderation for the same variable
(education) using another dataset (data 2) is carried out in Table 13.5. The test of
hypothesis is expected to produce the same result.
325
.92 .27 .29 .14
e1 e3 e4 e5
1 1 1 1
1.73 1.61
1.93
.79
1.00
e14
X1 1
.12
Y1 e10
1 1.00
1.00
.32
1
1.04 Y2 e11 .06
.34 Y 1.09
.17
1
Y3 e12
.17
1.08 1.21 .15
1
Y4 e13
1.07
X2 Fitness Indexes
1. ChiSq = 75.987
2. df = 50
.66 .79 1.00
.97 3. ChiSq/df = 1.520
4. GFI = .928
5. AGFI = .888
X24 X23 X22 X21 6. CFI = .985
7. RMSEA = .058
1 1 1 1
1.38 1.17 .12 .20
e9 e8 e7 e6
.40
Figure 13.11: Data 2: The Constrained Model. The Value of Chi-Square Is 75.987,
Df=50
326
.91 .27 .29 .14
e1 e3 e4 e5
1 1 1 1
1.53 1.41
1.70
.78
1.00
e14
X1 1
Y1 e10 .12
1 1.00
.70
.42
1
1.04 Y2 e11 .06
.39 Y 1.10
.17
1
Y3 e12
.23
1.08 1.21 .15
1
Y4 e13
1.07
X2 Fitness Indexes
1. ChiSq = 73.575
2. df = 49
.66 .79 1.00
.98 3. ChiSq/df = 1.502
4. GFI = .930
5. AGFI = .888
X24 X23 X22 X21 6. CFI = .985
7. RMSEA = .057
1 1 1 1
1.38 1.17 .12 .20
e9 e8 e7 e6
.40
327
Table 13.5: The Measures Required For Performing Moderation Test Using Data 2
The test of hypothesis for moderation found that the moderator variable
“respondents’ education” does not moderate the causal effects of X1 on Y. The results
of hypothesis testing are consistent for both datasets.
Example: Suppose that the researcher is also interested to determine whether the
same moderator variable (respondents’ education) moderate the causal effect of X2 on
Y – the same moderator variable but testing on different path. The same procedure that
has been executed earlier is applied.
328
e1 e3 e4 e5
1 1 1 1
e14
X1 1
Y1 e10
1 1
1
Y2 e11
Y
1
Y3 e12
1
1
Y4 e13
X2
1
e9 e8 e7 e6
constrained
Figure 13.13: The Constrained Model: The Coefficient in the Path of Interest (X2 to
Y) Where the Effect of Moderator Variable to Be Examined Is
Constrained To 1.
Recall: The parameter constraint is fixed in the path where the moderation effect
will be examined, and the data is split equally based on median of the moderator
variable. In the above example, the path of interest is from X2 to Y and the moderator
variable to be tested is respondents’ education.
The output for the constrained and unconstrained models is presented in Figure
13.14 and Figure 13.15 respectively. The test of moderation is carried out in Table 13.6.
329
.87 .18 .26 .24
e1 e3 e4 e5
1 1 1 1
1.57
1.00 1.44
1.21
1.56
e14 .09
X1 1
Y1 e10
1 1.00
-.10
.51 .05
1
.99 Y2 e11
.46 Y 1.06
.20
1
Y3 e12
1.00 1.02
1.12 .73 .12
1
Y4 e13
X2 Fitness Indexes
1. ChiSq = 185.363
1.00 2. df = 50
.69 .91 1.08 3. ChiSq/df = 3.707
4. GFI = .848
5. AGFI = .763
X24 X23 X22 X21 6. CFI = .930
7. RMSEA = .131
1 1 1 1
1.39 .94 .06 .20
e9 e8 e7 e6
.49
Figure 13.14: Data 1: The Constrained Model. The Value of Chi-Square Is 185.363,
Df=50
330
.86 .19 .24 .25
e1 e3 e4 e5
1 1 1 1
1.55
1.00 1.44
.72
1.54
e14 .09
X1 1
Y1 e10
1 1.00
.69
.51 .04
1
1.03 Y2 e11
.49 Y 1.09
.20
1
Y3 e12
.14 1.06
1.31 .72 .12
1
Y4 e13
X2 Fitness Indexes
1. ChiSq = 104.520
1.00 2. df = 49
.62 .83 1.01 3. ChiSq/df = 2.133
4. GFI = .906
5. AGFI = .851
X24 X23 X22 X21 6. CFI = .971
7. RMSEA = .085
1 1 1 1
1.41 .96 .05 .18
e9 e8 e7 e6
.51
331
Table 13.6: The measures required for performing moderation test using
Data 1
The test of hypothesis for moderation found that the moderator variable
“education” moderates the causal effects of X2 on Y.
The procedure for performing the test of moderation for the same variable
(education) using another dataset (data 2) is carried out in Table 13.7. The test of
hypothesis is expected to produce the same result that the respondents’ level of
education does moderate the causal effects of X2 on Y.
332
.91 .26 .30 .14
e1 e3 e4 e5
1 1 1 1
1.53 1.40
1.69
1.04
1.00
e14
X1 1
.12
Y1 e10
1 1.00
.06
.42
1
1.00 Y2 e11 .06
.36 Y 1.06
.18
1
Y3 e12
1.00
.85 1.18 .15
1
Y4 e13
1.06
X2 Fitness Indexes
1. ChiSq = 123.410
2. df = 50
.75 .90 1.00
1.10 3. ChiSq/df = 2.468
4. GFI = .884
5. AGFI = .819
X24 X23 X22 X21 6. CFI = .957
7. RMSEA = .097
1 1 1 1
1.36 1.15 .11 .26
e9 e8 e7 e6
.39
Figure 13.16: Data 2: The Constrained Model. The Value of Chi-Square Is 123.410,
Df=50
333
.91 .27 .29 .14
e1 e3 e4 e5
1 1 1 1
1.53 1.41
1.70
.78
1.00
e14
X1 1
.12
Y1 e10
1 1.00
.70
.42
1
1.04 Y2 e11 .06
.39 Y 1.10
.17
1
Y3 e12
.23
1.08 1.21 .15
1
Y4 e13
1.07
X2 Fitness Indexes
1. ChiSq = 73.575
2. df = 49
.66 .79 1.00
.98 3. ChiSq/df = 1.502
4. GFI = .930
5. AGFI = .888
X24 X23 X22 X21 6. CFI = .985
7. RMSEA = .057
1 1 1 1
1.38 1.17 .12 .20
e9 e8 e7 e6
.40
334
Table 13.7: The measures required for performing moderation test using Data 2
***The test is significant since the Chi-Square difference between the constrained
and unconstrained model is greater than 3.84. Recall: The Chi-Square value with 1
degree of freedom is 3.84.
Referring to Table 13.7: All fitness indexes for the unconstrained model is
significantly better (smaller Chi-Square) than the constrained model, indicating that the
two group’s coefficient differ. The results show strong support for the hypothesis that
education moderates the relationship between latent exogenous construct X 2 and its
corresponding latent endogenous construct Y.
To address this particular research question, the researcher needs to obtain the
standardized parameter estimate for the coefficient linking X2 and Y for both datasets.
The result is presented in Figure 13.18 for low education group, and Figure 13.19 for
high education group.
335
COMPARING THE GROUP EFFECT FOR MODERATOR VARIABLE
Suppose one has the following research question to address: In which group (low
education or higher education), the effect of moderator variable (education) is more
pronounced? To address this RQ, the researcher needs to obtain the standardized
estimate of the path of interest for both datasets. The procedure is demonstrated in
Figure 13.18 and Figure 13.19.
e1 e3 e4 e5
.93
.61 .90
.91
e14 .92
X1
Y1 e10
.48 .96
.97
.98 Y2 e11
.60 Y .93
.86
Y3 e12
.15 .79
.34 .63 .31
Y4 e13
X2 Fitness Indexes
1. ChiSq = 104.520
.94 2. df = 49 Standardized
.51 .70 .98 3. ChiSq/df = 2.133
.26 .49 .97 .88 4. GFI = .906
Slope
5. AGFI = .851
X24 X23 X22 X21 6. CFI = .971
7. RMSEA = .085
e9 e8 e7 e6
.44
Figure 13.18: The Standardized Beta for Low Education in Path X2 to Y (0.15)
336
The effect of X2 on Y is not significant for “low Education” group
e1 e3 e4 e5
.88 .86
.56 .95
e14 .90
X1
Y1 e10
.42 .95
.96
.98 Y2 e11
.58 Y .94
.89
Y3 e12
.22 .34 .62
.79 .35
Y4 e13
X2 Fitness Indexes
1.ChiSq = 73.575
.50 2.df = 49 Standardized
.61 .95 .92 3.ChiSq/df = 1.502
4.GFI = .930
Slope
.25 .37 .89 .84
5.AGFI = .888
X24 X23 X22 X21 6.CFI = .985
7.RMSEA = .057
e9 e8 e7 e6
.32
Figure 13.19: The Standardized Beta for High Education in Path X2 to Y (0.22)
337
The effect of X2 on Y is significant for “High Education” group
***Since the estimated beta coefficient or the slope measuring the causal effect of
X2 on Y differs significantly between the two groups namely low education and high
education, the researcher could conclude that “education” moderates the relationship
between X2 and Y.
In this case, the effect of “high education group” (slope 0.22) is more pronounced
compared to “low education group” (slope 0.15)
***Remember- when the two slopes differ, the two linear equations would interact
with each other at some point (interaction means moderation). The types of moderation
have is full moderation since one of the slope (slope 0.15) is not significant in testing
of hypotheses while the other slope (slope 0.22) is significant.
Note: If both groups (low education and high education) are significant then the
partial moderation occurs.
338
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
1 C
2 A
3 D
4 C
5 B
6 D
7 A
8 B
9 D
Essay Questions
4
i Causal
ii Descriptive
iii Descriptive
iv Descriptive
8 Descriptive study. Scope is the woman’s dress boutique within 40 km radius.
339
Suggested Solution for the Selected Exercises in the Chapter
340
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 D
2 C
3 A
4 A
5 C
6 D
7 D
8 A
9 C
10 D
11 B
12 B
13 D
14 D
15 A
16 A
17 D
18 A
19 D
20 B
341
Suggested Solution for the Selected Exercises in the Chapter
Essay Questions
10
a. Employees are not motivated to work. The possible causes of no motivation
include no salary increment given and also they are not being rewarded
accordingly by the management.
Dependent variable: Employees motivation level
Independent variables: Gender, Job Position, Amount of Salary, Length of
Service, Recognition, and Awards.
b. Research Questions:
i) What is the employees’ level of motivation?
ii) Does motivation increase with the increase in salary?
iii) What is the relationship between length of service and employees’
motivation level?
iv) Is there a difference in mean motivational level between male and female
employees?
v) Is motivation level greater among those who have served longer than
those who have served in a short period of time?
c. Objectives of the study:
i) To determine the level of motivation among employees.
Analysis: Frequencies and percentages of high, medium and low
motivation level.
ii) To determine the differences in mean motivational level between
male and female employees.
Analysis: t-test for two independent groups.
iii) To determine the relationship between length of service and
employees’ motivation level.
Analysis: Cross-tabulation and Chi-square test of association
d. Demographic profile:
Gender: Male ( ) Female ( )
Salary (per month) : RM________
Highest level of education: __________
Length of service: ___________
342
Suggested Solution for the Selected Exercises in the Chapter
1 a. Chapter 1
Introduction which include the background of study and organization in which
the research took place (if any)
State research objectives, research questions and hypotheses
Scope of study
Significance of study
Limitation of study
Chapter 2
Literature Review
Chapter 3
Research Methodology
- Research Design
- Sampling Design
- Instrument
- Process of Data Collection
- Types of Analyses
343
Suggested Solution for the Selected Exercises in the Chapter
1 c. What are the factors that may influence corporate officers to purchase the
magazine? What is perception of corporate officers towards the magazine?
2. Identify the variables of interest in the study. State how the study believes
these variables should relate to one another. This should be done based on
the theorized relationships among the variables.
Elaborate the nature and direction of the relationships that can be theorized
from the findings of previous research.
Provide a clear explanation of why the study would expect these
relationships to exist. The arguments could be drawn from the previous
research findings.
Draw a schematic diagram of the theoretical framework so that the reader
can see and easily comprehend the theorized relationships.
344
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 B
2 A
3 B
4 D
5 A
6 A
7 C
8 D
9 C
10 C
11 D
12 B
13 C
14 B
15 A
16 B
17 D
345
Suggested Solution for the Selected Exercises in the Chapter
Essay Questions
1
a Total students (population) = 12000
Total first year students (subpopulation) = 4000
Total sample (required) = 3000
Thus, the proportion to be sampled from first year = (4000/12000) x 300 = 100
b Both simple random sampling and systematic sampling method can be used. This
is due to the availability of sampling frame (list of students) and also the
characteristic within subpopulation is homogeneous.
c First of all, select six random numbers either through row or column arrangements.
Supposed the selected numbers are as shown in row 1 below. The amendment for
each number can be made as follows:
Take the first one digit from the first number, take the first two digits from the
second number, take the first three digits from the third number, and take all four
digits from the fourth number. For the fifth number, we take the last three digits,
and the sixth number we take the last two digits.
The
9156 7425 6535 2478 2810 1795
Note: The amendment for each number depends on the creativity of the researcher since all
the numbers selected are already random.
d Another method can be made by generating random numbers through computer.
We can ask the computer to select randomly 70 numbers ranging from 1 to 2800.
Random 70 1-2800.
e The method of data collection for this study can be made through either face to
face interview with the selected students or by giving them the self-administered
questionnaire personally to complete.
8 a) Method 1: Using personal interview where researcher meet and interview the
selected students at their respective faculties in the campus using questionnaire
that contain mostly structured questions with several unstructured questions.
346
Suggested Solution for the Selected Exercises in the Chapter
Administrative error
- sample selection error
interviewer error
347
Suggested Solution for the Selected Exercises in the Chapter
Advantages
- Cost less than other data collection method.
- Do not need many workers or people involve in the data collection and will
not waste the researcher time rather than go and meet them at their house.
Disadvantages
- The possible respondent may not get the mail or he or she did not want to
answer the questionnaire.
- Take a lot of time to be collected back to the researcher.
d) Questionnaire
- Let the respondent to answer the questions and it obtains the simplest
answer or data.
348
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 B
2 A
3 A
4 D
5 A
6 D
7 B
8 A
9 D
10 C
Essay Questions
1
a It would determine the method of data analysis
b i) Qualitative – ordinal scale
ii) Qualitative – ordinal scale
iii) Quantitative – Interval scale
2 a) Primary data
b) All repair workshops in Kota Bharu, Kelantan
c) The list of all repair workshops in Kota Bharu. The list could be obtained from
the office of Kota Bharu City Council
e) Personal meeting with the respondents to explain the purpose of the study and
obtain their consent before giving them the self-administered questionnaire to
answer at their convenient time. This method would increase the rate of response
and also the accuracy of information.
349
Suggested Solution for the Selected Exercises in the Chapter
18. Nominal: A scale in which the numbers or letters assigned to certain objects as
labels for identification or classification in a dataset.
Example 1 = Male, 2 = Female.
Ordinal: A scale that arranges objects or alternatives according to their
magnitudes – ordered arrangement.
Ratings of car service: Excellent, Good, Fair or Poor.
Interval: A scale that arranges objects or alternatives according to their
magnitudes; distinguishes ordered arrangement in units of equal intervals.
Measurement of psychological attributes such as feelings, motivation, etc.
Not happy Moderately happy Very happy
-1 0 1
The researcher can comment on the magnitude of differences but cannot
determine the actual strength or attitudes toward an object.
Temperature: -10ºC 0ºC 10ºC 20ºC
It cannot be said that 10ºC is twice as cold as 20ºC. 0ºC represents the lack of
temperature, but a relative point on the Celsius scale. It is a meaningful zero
(not absolute).
Ratio: A scale having absolute rather than relative quantities and possessing an
absolute zero, where there is an absence of a given attribute.
Money and weight possess an absolute zero and interval properties.
19 a) Nominal, ratio, nominal, nominal, ratio, ratio
b) Pie/simple bar, histogram/stem and leaf/ box plot, Pie/simple bar, Pie/simple
bar, histogram/stem and leaf/ box plot, histogram/stem and leaf/ box plot.
350
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 D
2 B
3 C
4 B
5 B
6 D
7 D
8 A
9 D
10 D
Essay Questions
351
Suggested Solution for the Selected Exercises in the Chapter
3. Cross-sectional: A study in which data are collected at one point in time such as a
public opinion survey, experimental study & case study. It is also known as a snap-
shot research.
Longitudinal: A study in which data are collected at more than one point in time. It
may consist of a series of cross-sectional surveys conducted with different people,
but asking the same or similar questions or it may be conducted with the same
people who are repeat-interviewed or surveyed.
4.
Factors to consider when choosing between observation and survey methods.
Observation Survey
- Purpose of Study (sensing the - Purpose of study when the
phenomena) study involved measuring social
- When the study does not involve phenomena descriptively using
interviewing respondents but a certain developed instrument;
instead requires direct ability to reach the target
observation of the event sample simultaneously.
occurring such as observing - Control of data collection
students’ interactions in the - Speed of data collection
classroom. - Time factor
- Control of data collection - Cost
- Time factor
- Cost
352
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 B
2 D
3 A
4 C
5 D
6 C
7 A
8 D
353
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 D
2 D
3 B
4 A
5 D
6 D
7 A
8 B
9 A
10 D
11 D
12 A
13 A
14 C
15 C
16 D
17 A
18 C
19 A
20 D
354
Suggested Solution for the Selected Exercises in the Chapter
Essay Questions
You have been selected as a respondent in this survey. Your sincere cooperation is
needed in providing the required information for our association. In this survey, I
would like to obtain information regarding your achievements, current activities and
future aspirations. Please read carefully all instructions pertaining to every section
and answer every question sincerely. All responses provided will be used to furnish
the database of our association. Thank you very much for your cooperation.
Sincerely,
………………………
Dr Zainudin Hj Awang (President)
b Should avoid being biased towards their political affiliation, types of their social
activities and their future aspirations which might be sensitive in certain extent.
5 Five common mistakes in developing a questionnaire (Any)
Avoid the use of complex language
Avoid Leading questions
Avoid loaded questions
Avoid double-barrel questions
Avoid burdensome questions
Avoid asking sensitive questions
355
Suggested Solution for the Selected Exercises in the Chapter
(ii) How satisfied are you with the car rental service?
Very dissatisfied Dissatisfied Neutral Satisfied Very satisfied
1 2 3 4 5
(iii)
Which brand of hand phone would you prefer to buy?(Tick √ one only)
Nokia □1
Samsung □2
Motorola □3
Sony Ericsson □4
8
Structured questions Unstructured questions
Advantage:
Advantage:
- It allows quick response from the
- It enables the respondents to
respondents
express general attitudes and
- Less time to administer the
opinions
questions
- Less biasing influence on
- Coding of responses is cheaper
response
and less time consuming
Disadvantage:
Disadvantage:
- Potential for interviewer bias is
- Respondents have less freedom
high
of expressing general attitudes
- Coding of responses is costly and
- More biasing influence on
time consuming.
response
356
Suggested Solution for the Selected Exercises in the Chapter
True-False Questions
true false
1 x
2 x
3 x
4 x
5 x
6 x
7 x
8 x
9 x
10 x
1 A
2 C
3 D
4 D
5 B
6 B
7 B
8 B
9 D
10 C
357
Suggested Solution for the Selected Exercises in the Chapter
Coding
- assign relevant codes (numeric or string) to the categories of responses
- label the codes according to the questions
assign different codes for missing values
358
Suggested Solution for the Selected Exercises in the Chapter
1)
a. Disagree with the process of preparing the report because the researcher should
begin the analysis with the basic descriptive analysis of each question or variable
of interest. It should then be followed by some bivariate descriptive analysis and
inferences. This includes conducting hypotheses testing, such as t-test and chi-
square test of independence.
b. Nominal and Ordinal.
c. From a total of 2,715 respondents, it was found that 1,653 did not respond to the
question and hence were excluded from the analysis. Thus valid percentage will
be reported.
The table shows that majority of the respondents, i.e. 62.4% (650 out of 1,653)
indicated that their work is interesting. This is followed by 29.1% (303) of
respondents who were slightly less favorable on the statement.
From the ANOVA table, since F-statistics = 11.983 and p-value = 0.001 ˂ 0.05,
therefore Ho is rejected. This concludes that at least one mean sale figures is
different.
c) Based on the post hoc multiple comparisons, the following pairs of policies are
found to be significantly different in their mean sale figures:
Policy 1 and Policy 3 (p = .001)
Policy 2 and Policy 3 (p = .009)
3 H0: The nutritional program is not effective in reducing blood cholesterol level
(μA ≥ μB)
H1: The nutritional program is effective in reducing blood cholesterol level (μA ≤ μB)
From the Wilcoxon signed ranks test, Z-test = -2.105 and p-value = 0.035 ˂ 0.05,
359
Suggested Solution for the Selected Exercises in the Chapter
Front location has the highest maximum and minimum sales volume at 8.6 and
4.0, respectively. Rear location gives the second highest max and min sales
volume at 6.0 and 2.2, respectively, followed by middle location with the least max
and min sales volume.
5 Disagree with the process of preparing the report because the researcher should
begin the analysis with the basic descriptive analysis of each question or variable
of interest. It should then be followed by some bivariate descriptive analysis and
inferences. This includes conducting hypotheses testing, such as t-test and chi-
square test of independence.
6 Nominal and Ordinal.
7 Majority is males (54%) and stays off-campus (51%) with friends. The age of the
respondents ranges from 13 (it’s a mistake here but too late to change) to 27
years old with an average of 20.72 years and a standard deviation of 2.656 years.
Slightly less than 20 per cent (19%) smokes regularly while a higher percentage
(21%) smokes occasionally
8 The table indicates that there is a higher proportion among male students who
smoke regularly (29.6%) or occasionally (22.2%) compared with female students
(5.5% and 19.6%, respectively). This observation is confirmed by the results of
the Chi-square test of association which shows that there is a significant
difference in the smoking habit between males and females at the 5% level of
significance (p-value=0.007˂0.05)
9 a) Target population: All employees working at HyPetrol Sdn Bhd.
Sample: Representatives of local and foreign workers
Sampling frame can be acquired from the Manager of the Human Resource Dept
of the company.
Problem Statement: There is a concern for low productivity in the manufacturing
b) of oil gaskets for automobile and small tractor engine. Based on informal
information, the management found that the outcome of low productivity relates to
employees’ dissatisfaction with the new company policy.
c) A suitable research design is descriptive research via survey method.
Purpose of descriptive research is to describe characteristics of a population or a
360
Suggested Solution for the Selected Exercises in the Chapter
phenomenon and to determine the answers to who, what, when, where, and how
questions. This is done through survey method where it is used to obtain
information from respondents either through personal interview, telephone, mail,
and other suitable methods.
d)
Dependent – Job Satisfaction
Independent – Age, length of service, gender, opportunity for advancement and
origin.
Independent variables
Age Dependent variable
Length of service Job Satisfaction
Gender
Opportunity for advancement
Origin (local for foreign)
There is a difference in mean job satisfaction between the local and foreign
e) workers
- There is a relationship between age group and job satisfaction level of workers
- Foreign workers are more satisfied with their job than the local workers
- There is a difference in mean job satisfaction between different age groups.
Or any other appropriate hypotheses
361
Index
captive audience, 93
categorical scale, 117, 118
causal factors, 41
causal research design, 40, 41
cause and effect relationship, 40
CD-ROM database, 158
census, 45, 84
Central Limit Theorem, 85
characteristics of interest, 5, 94, 110
Chi-Square test for correlation, 49, 115
classification information, 179
close-ended questions, 171
cluster sampling, 92
coding, 76 , 217
coefficient of determination, 249
coefficient of variation, 115
cohort analysis, 43
comparative scale, 117
comparison scale (ranking scale), 122
complete mediation, 306
complete moderation, 313
computerized database, 157
computerized search, 158
conclusive evidence, 37, 39, 43
conference proceedings, 28
confidentiality, 13
confirm coding, 216
confirmatory factor analysis, 268, 276
constant sum scaling, 124
constrained model, 320, 327
construct of primary interest, 69
construct validity, 277
construct reliability,
content in a questionnaire, 169, 172
continuous variable, 107
convenience sampling, 93, 94
corporate image, 9, 24, 70
corrective action, 2
correlational study, 40, 271
confirmatory factor analysis (CFA), 269
constrained model, 319, 320, 323
construct validity, 276
covariance, 252, 301
convergent validity, 276
credible scientific finding, 11
critical ratio, 299
critically and objectively, 3
Cronbach’s alpha, 234
cross sectional design, 43, 77
customer loyalty, 9, 61, 110
customer satisfaction, 3, 24, 61
editing, 76
eigenvalue, 230
endogenous construct, 272
error in the equation, 247
error in the measurement, 248
establish rapport, 179
existing phenomenon, 38
exogenous constructs, 252, 273, , 277, 297
experimental research, 40
explicit directions,181
exploratory research, 37
external sources, 155
heterogeneous population, 90
historical information, 9, 104, 153
homogeneity, 72
homogeneous characteristics, 72, 88, 90
hypothesis,10, 235
hypothesized link, 255
hypothetical constructs, 244
I
job satisfaction, 9
judgmental sampling, 94
sample size, 85
sample, 29
sampling adequacy, 229
sampling and data collection procedure, 73
sampling design, 82
sampling distribution of the mean, 85
sampling distribution, 75
sampling error, 45, 72, 84
sampling frame, 73, 81, 88, 92
sampling method, 72
sampling methodology, 37
sampling procedure, 88
sampling technique, 80, 81
sampling units, 85, 95,
schematic diagram, 241
scientific method, 2, 38, 85
scope of a research project, 12, 77, 83
screening alternatives, 37
secondary data, 103, 153
secondary research, 104
second generation technique, 268
second-hand data, 104
selection bias, 94
self-addressed envelope, 74
self-administered questionnaires, 44, 109
SEM, 268
semantic differential scale, 120
semantic meaning, 120
sense of authority, 182
sensitive question, 180
service quality, 24, 61, 62, 70
significance of the research, 57
significance of the study, 66
significance value, 230
significant and direct influence, 242, 257
simple random sampling, 88, 89
sincere response, 113
skewness, 291
snowball sampling, 95
source databases, 157
specific question, 180
squared multiple correlation, 297
standard deviation, 47, 224, 249
standard error, 47, 224
standardized estimate, 294, 337
standardized model, 249
stapel scale, 121
statement of problem, 59
statistical analysis procedure, 49
statistical analysis techniques, 84, 244
statistical information, 3
statistical measures,47
statistical reliability, 72
statistical significance, 85
statistical test, 12, 75
statistical validity, 72
statistically invalid, 84
strata, 90
stratified sampling method, 72, 83
Structural Equation Modelling (SEM), 49, 245, 292
structure of questionnaire, 170
structure of response, 169
structure to the interview, 167
structured questionnaire, 44, 93, 103
structured questions, 167
subgroups in a population, 43, 94
symptom to a problem, 18, 19
systematic procedure, 210
systematic random sampling, 83
systematic sampling technique, 89
systematic search, 1
T
unbiased, 72
unconstrained model, 321, 324, 327
underlying problem, 3, 25, 48
underlying theory, 159
unidimensionality, 276
unit of analysis, 20
univariate associational question, 49
univariate descriptive question, 49
univariate difference question, 49
unstandardized regression weights, 298
unstructured questions, 171
Z-test, 235
Bibliography
Zainudin, A. and Zaihan, O. (2004). The Impact of Corporate Image on Perceived Quality in
Healthcare Industry. Proceeding of Conference of Social Science Research (CSSR
2005). Institute of Research Development and Commercialization UiTM,
Zainudin, A., Rustam, S. and Farouk, M.Z. (2005). Service quality and customer satisfaction
in the environment of higher education: A study on the public and private university
students in Malaysia, Institute Research Development and Commercialization, IRDC,
UiTM.
Zainudin, A., W. Fauzi, W.M. and Azuhari, C.M. (2005b). Achieving Students’ Satisfaction
in Higher Education: Closing the gaps. International Conference on Quantitative
Sciences and its Application (ICOQSIA 2005). Faculty of Quantitative Sciences,
UUM.
Zainudin, A. (2007). The Effects of Service Quality and Corporate image on Students’
Loyalty in Postgraduate Education: The case study of a local university, Unpublished
Thesis, Universiti Malaysia Terengganu (UMT), Malaysia.
Zainudin, A, Azahar, Y, and Masayu, F (2008). The influence of corporate reputation on
competitive advantage to the firms in household furniture industry. Paper presented at
The International Furniture Conference IFCx, PWTC Kuala Lumpur 28-30 July
2008.
Anderson, J.C., & Gerbing, D.W. (1988).Structural Equation Modeling in Practice: A review
of recommended two steps approach. Psychological Bulletin, 103(3), 411-423.
Arbuckle, J.L. (2005). AMOS 6.0 User’s Guide. Spring House PA: Amos Development
Corporation.
Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation models. The
Academy of Marketing Science. 16(1), 74-94.
Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable distinction in social
psychological research: Conceptual, strategic and statistical considerations. Journal of
Personality and Social Psychology, 51: 1173-1182.
Bentler, P. & Chou, C. (1987). Practical Issues in Structural Equation Modeling. Sociological
Methods and Research. 69(16): 78-117.
Bentler, P.M. & Bonett, D.G. (1980). Significance tests and goodness of fit in the analysis of
covariance structures. Psychological Bulletin. 88: 588–606.
Bentler, P.M. (1990). Comparative fit indexes in structural models. Psychological Bulletin
107: 238–246
Bollen, K.A. (1986). Sample size and Bentler and Bonett’s non-normed fit index. Psychometrika.
51:375–377.
Bollen, K.A. (1989b). A new incremental fit index for general structural equation models.
Sociological Methods and Research 17: 303–316.
Byrne, B.M. (2001). Structural Equation Modeling with Amos: Basic Concepts, Applications,
and Programming. Mahwah, NJ: Erlbaum.
Byrne, B.M. (2010). Structural Equation Modeling With AMOS: Basic Concepts,
Applications, and Programming 2nd Edition. Routledge. New York.
Cronbach. L.J. (1951). Coefficient Alpha and the Internal Structural of Tests. Pychometrica.
16(3): 297-334
Edwards, J.R., & Lambert L. S. (2007). Methods for integrating moderation and mediation: A
general analytical framework using moderated path analysis. Psychological Methods,
12: 1-22.
Frazier, P.A., Tix, A. P. & Barron, K. E. (2004). Testing moderator and mediator effects in
counseling psychology research. Journal of Counseling Psychology, 51: 115-134.
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., & Tatham, R.L. (2006). Multivariate
data analysis. New Jersey: Pearson Prentice Hall.
Hair, J.F., Black, W.C., Rabin, B.J. & Anderson, R.E. (2010). Multivariate Data Analysis (7th
Ed). Englewood Cliffs, NJ: Prentice Hall.
Hair, J.F.,Anderson, R.E., Tatham, R.L., & Black, W.C. (1995). Multivariate Data Analysis
with Readings (4th Ed). Englewood Cliffs, NJ: Prentice Hall.
Hayes, A.F., & Matthes, J. (2009). Computational procedures for probing interactions in OLS
and logistic regression: SPSS and SAS implementations. Behavior Research Methods,
41: 924-936.
Holmes-Smith, P., Coote, L., & Cunningham, E. (2006). Structural Equation Modeling:
From the Fundamental to Advanced Topics. Melbourne: Streams.
Jöreskog, K.G. & Sörbom, D. (1984). LISREL-VI user’s guide (3rd ed.). Mooresville, IN:
Scientific Software.
Kenny, D.A., & Judd, C.M. (1984). Estimating the nonlinear and interactive effects of latent
variables. Psychological Bulletin, 96: 201-210.
Klein, A.G., & Moosbrugger H. (2000). Maximum likelihood estimation of latent interaction
effects with the LMS method. Psychometrika, 65: 457-474.
Kline, R.B. (2005). Principles and Practice of Structural Equation Modeling (2nd ed.).New
York: The Guilford Press.
Kraemer, H.C., Wilson G.T., Fairburn, C.G., & Agras, W.S. (2002). Mediators and
moderators of treatment effects in randomized clinical trials. Archives of General
Psychiatry, 59: 877-883.
Kraemer, H.C., Stice, E., Kazdin, A., Offord, D., & Kupfer, D. (2001). How do risk factors
work together? Moderators, mediators, independent, overlapping, and proxy risk
factors. American Journal of Psychiatry, 158: 848-856.
Mahadzirah, M. & Zainudin, A. (2009). Building corporate image and securing students’
loyalty in the Malaysian higher learning industry. Journal of International Management
Studies. Vol. 4 (1): 30-40.
Muller, D., Judd, C. M., & Yzerbyt, V. Y. (2005). When moderation is mediated and
mediation is moderated. Journal of Personality and Social Psychology, 89: 852-863.
Norsalihah, A.R., Zainudin, A., and Nor Hazreeni, H. (2010). The effects of information and
communication technology on security compliance. Journal of Statistical Modeling and
Analytics. Vol 1 (1): 28-44.
Ping, R.A. (1996). Latent variable interaction and quadratic effect estimation: A two-step
technique using structural equation analysis. Psychological Bulletin, 119, 166-175.
Shammout, A.B. (2007). Evaluating Extended Relationship MarketingModel for Arab Guests
of Five Star Hotels. Unpublished Thesis: Faculty of Business and Law, Victoria
University Melbourne.
Tanaka, J.S. & Huba, G.J. (1985). A fit index for covariance structure models under arbitrary
GLS estimation. British Journal of Mathematical and Statistical Psychology 38: 197–
201.
Zainudin, A., Junaidah, H.A., & Nazmi, M.Z. (2010). Modeling job satisfaction and work
commitment among lecturers. Journal of Statistical Modeling and Analytics. Vol 1 (22):
45-59
Zainudin, A. (2010). Research Methodology for Business and Social Science. Shah Alam:
Universiti Teknologi MARA Publication Centre (UPENA).
Zainudin, A. (2007). Modeling the Loyalty of Outgoing Undergraduates towards
Postgraduate Education at their University. Thesis: Faculty of Management and
Economics, Universiti Malaysia Terengganu.