Professional Documents
Culture Documents
346EN150
1 – 24
ANNAMALAI UNIVERSITY
DIRECTORATE OF DISTANCE EDUCATION
F
RESEARCH METHODOLOGY AND STATISTICS
LESSONS: 1 – 24
Copyright Reserved
(For Private Circulation Only)
2
Editorial Board
Members
Dr. K. Rajendran
Dean
Faculty of Arts
Annamalai University
Annamalainagar.
Dr. A. Rajamohan
Dr. C. Samudhrarajakumar
Professor and Head
Professor and Head Management Wing, D.D.E.
Department of Business Administration Annamalai University
Annamalai University Annamalainagar.
Annamalainagar.
Internals
Dr. K.Anandanatarajan Dr. K.Tamizh Jyothi
Assistant Professor Assistant Professor
Management Wing, D.D.E Management Wing, D.D.E
Annamalai University Annamalai University
Annamalainagar. Annamalainagar.
Externals
Dr. P.Vikkraman Dr. C. Vethirajan
Director i/c Professor of corporate Secreteriship
Department of Management Studies Alagappa University
Anna University Karaikudi
Coimbatore
Lesson Writers
Units: I – III Units: IV – VI
Dr. S. Arul Kumar Dr. N. Kathirvel
Assistant Professor Assistant Professor
Management Wing, D.D.E Department of Commerce
Annamalai University
Goverenment Arts College
Annamalainagar.
Udumalpet
3
SYLLABUS
Objective
To equip the students with the basic understanding of the research
methodology and provide an insight into the application of modern analytical tools
and statistical techniques for the purpose of management decision making.
Unit–I
Research – Meaning – Types – Nature and scope of research – Review of
Literature - Problem formulation – Statement of research Objective – Value and cost
of information – Research Questions – Research Gap - Decision theory –Research
process – Research designs – Exploratory – Descriptive – Experimental Research.
Unit–II
Methods of data collection – Observational and Survey methods – Field work
plan – Administration of surveys – Training the field investigators – Sampling
methods – Sample size – Source of Data – Primary – Secondary - Preparation of
Interview Schedule - Questionnaire Design - Attitude measurement techniques –
Scaling Techniques.
Unit–III
Testing of Hypothesis - Hypothesis testing: Chi-square test - one sample and
two samples tests for means and proportions of large samples (z-test), one sample
and two sample tests for means of small samples (t-test), Standard Deviations – F-
test (ANOVA) – Correlation – Regression.
Unit–IV
Non-Parametric Methods - Sign test for paired data - Rank sum test - Mann –
Whitney U test and Kruskal Wallis test - One sample run test - Chi-square tests for
independence of attributes and goodness of fit.
Unit V
Multivariate Analysis - Factor Analysis - MANOVA - Discriminant Analysis,
Structural Equation Modeling - Conjoint analysis – Multidimensional Scaling –
Cluster Analysis (Theoretical Applications Only). Role of Statistics Software in
Business Decision.
Unit–VI
Research Report – Structure of Research Report – Title of a Report – Table of
Contents – Synopsis – Introduction and Body of a Report – Tables and Diagrams -
Results and Discussions – Recommendations, Conclusion and Scope for Further
Research –References – Appendices.
4
References Books
1. Gupta S: Research Methodology and Statistical Techniques, New Delhi, Deep
& Deep Publication (P) Ltd., 2002.
2. Tripathi P.C.: A Textbook of Research Methodology, New Delhi, Sultan Chand
& Sons, 2002.
3. R. Cauvery, U.K. Sudhanayak, R.Meenakshi, Research Methodology,
S.Chand, Chennai, 2012.
4. Dr.D.R. Krishnaswamy and Dr.M.Ranganathan, Methodology of Research in
Social Sciences, Second Edition, Himalaya Publishing House, 2014.
5. Dr.B.G. Satyaprasad and Dr.O.R. Krishnaswamy, Business Research
Methods, Fifth Edition, 2010.
6. J.K. Sachdeva, Business Research Methodology, Second Edition, Himalaya
Publishing House, 2013.
Journals and Magazines
1. Journal of Management Research
2. International Journal of Social Science Research
3. Journal of Educational and Social Science Research
4. International Journal of Research in Humanities and Social Sciences
5. Advances in Social Science Research Journal.
Web Resources
1. www.uk.sagepub.com
2. wps.pearsoned.co.uk/ema_uk_he_saunders_resmethbus_6
3. global.oup.com/uk/orc/busecon/business/brymanbrm3e
4. highered.mcgraw-hill.com/sites/0073373702/student_view0
5. www.wiley.com/WileyCDA/.../productCd-111994225X.html
5
CONTENT
Lesson Title Page No.
No.
1 Basics of Research 1
2 Review of Literature 13
3 Research Process 29
4 Research Design 37
5 Sources of Data 43
6 Measurement and Scaling Techniques 66
7 The Nature of Field Work 80
8 Sampling 88
9 Hypotheis Testing 104
10 T-Test and Z-Test 137
11 F-Test (ANOVA) 151
12 Measure of Relationship – Correlation and 160
Regression
13 Non-Parametric Test 181
14 Sign Test 186
15 CHI Square Test 198
16 Mann-Whitney Test 208
17 Multivariate Statistical Analysis 218
18 Kruskal-Wallis Test 231
19 Structural Equation Modeling 239
20 Cluster Analysis 253
21 Research Report 263
22 Types of Report 269
23 Mechanics of writing Report 274
24 Precautions for writing Report 282
6
1
LESSON – 1
BASICS OF RESEARCH
1.1 INTRODUCTION
Research in simple terms, refers to a search for knowledge. It is also known as
a scientific and systematic search for information on particular topic or issue. It is
also known as the art of scientific investigation. Several social scientists have
defined research in different ways
In the Encyclopedia of Social Sciences, D. Slesinger and M. Stephension (1930)
defined research as "the manipulation of things, concept or symbols for the purpose
of generalizing to extend, correct or verify knowledge, whether that knowledge
aids in construction of theory or in practice of an art".
According to Redman and Mory (1923), defined research is a "systematized
effort to gain new knowledge". It is an academic activity and therefore the term
should be used in a technical sense. According to Clifford Woody (Ko thari 1988)
research comprises "defining and redefining problems, formulating hypothesis or
suggested solutions; collecting, organizing and evaluating data; making deductions
and reaching conclusions; and finally, carefully testing the conclusions to
determine whether they fit the formulating hypothesis".
Thus, research is an original addition to the available knowledge, which
contributes to its further advancement. It is an attempt to pursue truth through the
methods of study, observation, comparison and e xperiment. In sum, research is the
search for knowledge, using objective and systematic methods to find solution to a
problem.
1.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Meaning of research
Scope and Objectives of research
Criteria for good research
Types of research
1.3 CONTENTS
1.3.1 Objectives of Research
1.3.2 Motivation in Research
1.3.3 Significance of Research
1.3.4 Scope of Research
1.3.5 Criteria of Good Research
1.3.6 Types of Research
1.3.1 Objectives of Research
The objective of research is to discover answers to questions by applying
scientific procedures. In the other words, the main aim of research is to find out
2
truth which is hidden and has not yet been discovered. Although every research
study has its own specific objectives, research objectives may be broadly grouped as
follows:
To gain familiarity with or new insights into a phenomenon (i.e.,
formulative research studies);
To accurately portray the characteristics of a particular individual, group,
or a situation (i.e., descriptive research studies);
To analyse the frequency with which something occurs (i.e., diagnostic
research studies); and
To examine a hypothesis of a causal relationship between two variables
(i.e., hypothesis-testing research studies).
1.3.2 Motivation in Research
What makes people to undertake research? This is a question of fundamental
importance. The possible motives for doing research may be either one or more of
the following:
Desire to get a research degree along with its conseque ntial benefits;
Desire to face the challenge in solving the unsolved problems, i.e., concern
over practical problems initiates research;
Desire to get intellectual joy of doing some creative work;
Desire to be of service to society;
Desire to get respectability.
However, this is not an exhaustive list of factors motivating people to
undertake research studies. Many more factors such as directives of government,
employment conditions, curiosity about new things, desire to understand causal
relationships, social thinking and awakening, and the like may as well motivate (or
at times compel) people to perform research operations.
1.3.3 Significance of Research
"All progress is born of inquiry. Doubt is often better than overconfidence, for
it leads to inquiry, and inquiry leads to invention" is a famous Hudson Maxim in
context of which the significance of research can well be understood. Increased
amounts of research make progress possible. Research inculcates scientific and
inductive thinking and it promotes the development of logical habits of thinking and
organization.
The role of research in several fields of applied economics, whether related to
business or to the economy as a whole, has greatly increased in modern times. The
increasingly complex nature of business and government has focused attention on
the use of research in solving operational problems. Research, as an aid to
economic policy, has gained added importance, both for government and business.
Research provides the basis for nearly all government policies in our economic
system. For instance, government's budgets rest in part on an analysis of the needs
and desires of the people and on the availability of revenues to meet these needs.
3
The cost of needs has to be equated to probable revenues and this is a field where
research is most needed. Through research we can devise alternative policies and
can as well examine the consequences of each of these alternatives.
Decision-making may not be a part of research, but research certainly
facilitates the decisions of the policy maker. Government has also to chalk out
programmes for dealing with all facets of the country's existence and most of these
will be related directly or indirectly to economic conditions. The plight of cultivators,
the problems of big and small business and industry, working conditions, trade
union activities, the problems of distribution, even the size and nature of defense
services are matters requiring research. Thus, research is considered necessary
with regard to the allocation of nation's resources. Another area in government,
where research is necessary, is collecting information on the economic and social
structure of the nation. Such information indicates what is happening in the
economy and what changes are taking place. Colle cting such statistical information
is by no means a routine task, but it involves a variety of research problems. These
day nearly all governments maintain large staff of research technicians or experts
to carry on this work. Thus, in the context of govern ment, research as a tool to
economic policy has three distinct phases of operation, viz., (i) investigation of
economic structure through continual compilation of facts; (ii) diagnosis of events
that are taking place and the analysis of the forces underlyi ng them; and (iii) the
prognosis, i.e., the prediction of future developments.
Research has its special significance in solving various operational and
planning problems of business and industry. Operations research and market
research, along with motivational research, are considered crucial and their results
assist, in more than one way, in taking business decisions. Market research is the
investigation of the structure and development of a market for the purpose of
formulating efficient policies for purc hasing, production and sales. Operations
research refers to the application of mathematical, logical and analytical techniques
to the solution of business problems of cost minimisation or of profit maximisation
or what can be termed as optimisation problems. Motivational research of
determining why people behave as they do is mainly concerned with market
characteristics. In other words, it is concerned with the determination of
motivations underlying the consumer (market) behaviour. All these are of great help
to people in business and industry who are responsible for taking business
decisions. Research with regard to demand and market factors has great utility in
business. Given knowledge of future demand, it is generally not difficult for a firm,
or for an industry to adjust its supply schedule within the limits of its projected
capacity. Market analysis has become an integral tool of business policy these days.
Business budgeting, which ultimately results in a projected profit and loss account,
is based mainly an on sales estimate which in turn depends on business research.
Once sales forecasting is done, efficient production and investment programmes
can be set up around which are grouped the purchasing and financing plans.
Research, thus, replaces intuitive business decisions by more logical and scientific
decisions.
4
of which fall in the category of pure re search. As a matter of fact, the validity of
your findings entirely depends upon the soundness of the research methods and
procedures adopted by you.
Within any profession, where you directly or indirectly provide a service, such
as health (nursing, occupational therapy, physiotherapy, community health, health
promotion and public health), education, psychology or social work, the application
of research can be viewed from four different perspectives:
the service provider;
the service administrator, manager and/or planner;
the service consumer; and
the professional.
1.6 Criteria of Good Research
Whatever may be the types of research works and studies, one thing that is
important is that they all meet on the common ground of scientific method
employed by them. One expects scientific research to satisfy the following criteria:
The purpose of the research should be clearly defined and common
concepts be used.
The research procedure used should be described in sufficient detail to
permit another researcher to repeat the research for further advancement,
keeping the continuity of what has already been attained.
The procedural design of the research should be carefully planned to yield
results that are as objective as possible.
The researcher should report with complete frankness, flaws in procedural
design and estimate their effects upon the findings.
The analysis of data should be sufficiently adequate to reveal its
significance and the methods of analysis used should be appropriate. The
validity and reliability of the data should be checked carefully.
Conclusions should be confined to those justified by the data of the
research and limited to those for which the data provide an adequate
basis.
Greater confidence in research is warranted if the researcher is
experienced, has a good reputation in research and is a person of
integrity.
6
7
the extent of their utilisation. Types of service are the qualitative aspect of the study
as finding out about them entails description of the services. The extent of
utilisation of the services is the quantitative aspect as it involves estimating the
number of people who use the services and calculating other indicators that reflect
the extent of utilisation.
1.4 REVISION POINTS
1. Research is a "systematized effort to gain new knowledge
2. The objective of research is to discover answers to questions by applying
scientific procedures
3. Research provides the basis for nearly all government policies in our
economic system
4. Every profession uses research methods in varying amounts in many areas
5. Pure research involves developing and testing theories and hypotheses that
are intellectually challenging to the researcher but may or may not have
practical application at the present time or in the future.
6. A descriptive study attempts to describe systematically a situation, problem,
phenomenon, service or programme, or provides information about, say, the
living conditions of a community, or describes attitudes towards an issue
1.5 INTEXT QUESTIONS
1. What are the objectives of research?
2. State the significance of research.
3. What is the importance of knowing how to do research?
1.6 SUMMARY
This chapter outlined the Importance of research. Difference between basic
and applied research have been dealt in detail. This chapter has given the Meaning,
scope, types, structure of the Research.
1.7 TERMINAL EXERCISES
1. Research in simple terms, refers to a ---------------- for knowledge
2. The objective of research is to discover answers to questions by applying ----
------------- procedures
3. Research has its special significance in solving various operational and
planning problems of ----------------
1.8 SUPPLEMENTARY MATERIALS
1. Cherryholmes, C. H. (1992). Notes on pragmatism and scientific realism.
Educational Researcher,l 4, August-September, 13-1 7.
2. http://www.usg.edu/galileo/skills/unit01/infoage01_03.phtml
3. http://library.queensu.ca/research/subjects
12
1.9 ASSIGNMENTS
1. “Creative management, whether in public administration or private industry,
depends on methods of inquiry that maintain objectivity, clarity, accuracy
and consistency”. Discuss this statement and examine the significance of
research”.
2. “Research is much concerned with proper fact finding, analysis and
evaluation.” Do you agree with this statement? Give reasons in support of
your answer.
1.10 SUGGESTED READINGS/ REFERENCE BOOKS
1. Hollway, W. and Jefferson, T. (2000) Doing Qualitative Research Differently:
Free Association, Narrative and the Interview Method, London: Sage.
2. Schwandt, T. (1997) Qualitative Inquiry: A Dictionary of Terms, Thousand
Oaks, CA: Sage.
3. Kothari, C.R, 1985, Research Methodology- Methods and Techniques, New
Delhi, Wiley Eastern Limited.
4. Kumar, Ranjit, 2005, Research Methodology-A Step-by-Step Guide for
Beginners, (2nd.ed.), Singapore, Pearson Education.
1.11 LEARNING ACTIVITIES
1. Using your local and national newspapers compile a list of career
opportunities in the area of marketing research.
1.12 KEYWORDS
Research: It is one of the ways of finding answers to your professional and
practice questions. However, it is characterised by the use of tested procedures and
methods and an unbiased and objective attitude in the process of exploration
Systematic research: It means that research is structured with specified steps
to be taken in a specified sequence in accordance with the well defined set of rules.
Systematic characteristic of the research does not rule out creative thinking but it
certainly does reject the use of guessing and intuition in arriving at conclusions.
Pure research: In the social sciences Pure research involves developing and
testing theories and hypotheses that are intellectually challenging to the researcher
but may or may not have practical application at the present time or in the future.
Thus such work often involves the testing of hypotheses containing very abstract
and specialised concepts.
13
LESSON – 2
REVIEW OF LITERATURE
2.1 INTRODUCTION
One of the essential preliminary tasks when you undertake a research study is
to go through the existing literature in order to acquaint yourself with the available
body of knowledge in your area of i nterest. Reviewing the literature can be time
consuming, daunting and frustrating, but it is also rewarding. The literature review
is an integral part of the research process and makes a valuable contribution to
almost every operational step. It has value even before the first step; that is, when
you are merely thinking about a research question that you may want to find
answers to through your research journey. In the initial stages of research it helps
you to establish the theoretical roots of your study , clarify your ideas and develop
your research methodology. Later in the process, the literature review serves to
enhance and consolidate your own knowledge base and helps you to integrate your
findings with the existing body of knowledge. Since an importan t responsibility in
research is to compare your findings with those of others, it is here that the
literature review plays an extremely important role. During the write -up of your
report it helps you to integrate your findings with existing knowledge - that is, to
either support or contradict earlier research. The higher the academic level of your
research, the more important a thorough integration of your findings with existing
literature becomes. In summary, a literature review has the following functions:
It provides a theoretical background to your study.
It helps you establish the links between what you are proposing to
examine and what has already been studied.
It enables you to show how your findings have contributed to the existing
body of knowledge in your profession.
It helps you to integrate your research findings into the existing body of
knowledge.
In relation to your own study, the literature review can help in four ways. It
can:
Bring clarity and focus to your research problem;
Improve your research methodology;
Broaden your knowledge base in your research area; and
Contextualize your findings.
2.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Review of literature
Research problem
Research objectives
14
Research question
Research gap
2.3 CONTENTS
2.3.1 How to review the literature?
2.3.2 The Research problem
2.3.3 Objectives
2.3.4 Research questions
2.3.5 Research gap
Bringing clarity and focus to your research problem
The literature review involves a paradox. On the one hand, you cannot
effectively undertake a literature search without some idea of the problem you wish
to investigate. On the other hand, the literature review can play an extremely
important role in shaping your research problem because the process of review ing
the literature helps you to understand the subject area better and thus helps you to
conceptualise your research problem clearly and precisely and makes it more
relevant and pertinent to your field of enquiry. When reviewing the literature you
learn what aspects of your subject area have been examined by others, what they
have found out about these aspects, what gaps they have identified and what
suggestions they have made for further research. All these will help you gain a
greater insight into your own research questions and provide you with clarity and
focus which are central to a relevant and valid study. In addition, it will help you to
focus your study on areas where there are gaps in the existing body of knowledge,
thereby enhancing its relevance.
Improving your research methodology
Going through the literature acquaints you with the methodologies that have
been used by others to find answers to research questions similar to the one you
are investigating. A literature review tells you if others have used procedures and
methods similar to the ones that you are proposing, which procedures and methods
have worked well for them and what problems they have faced with them. By
becoming aware of any problems and pitfalls, you will be better positioned to select
a methodology that is capable of providing valid answers to your research question.
This will increase your confidence in the methodology you plan to use and will
equip you to defend its use.
Broadening your knowledge base in your research area
The most important function of the literature review is to ensure you read
widely around the subject area in which you intend to conduct your research study.
It is important that you know what other researchers have found in regard to the
same or similar questions, what theories have been put forward and what gaps
exist in the relevant body of knowledge. When you undertake a research project for
a higher degree (e.g. an MA or a PhD) you are expected to be an expert in your area
of research. A thorough literature review helps you to fulfil this expectation.
15
Another important reason for doing a literature review is that it helps you to
understand how the findings of your study fit into the existing body of knowledge.
Enabling you to contextualise your findings
Obtaining answers to your research questions is comparatively easy: the
difficult part is examining how your findings fit into the existing body of knowledge.
How do answers to your research questions compare with what others have found?
What contribution have you been able to make to the existing body of knowledge?
How are your findings different from those of others? Undertaking a literature
review will enable you to compare your findings with those of others and answer
these questions. It is important to place your findings in the context of what is
already known in your field of enquiry
2.3.1 How to review the literature?
If you do not have a specific research problem, you should review the literature
in your broad area of interest with the aim of gradually narrowing it down to what
you want to find out about. After that the literature review should be focused
around your research problem. There is a danger in reviewing the literature without
having a reasonably specific idea of what you want to study. It can condition your
thinking about your study and the methodology you might use, resulting in a less
innovative choice of research problem and methodology than otherwise would have
been the case. Hence, you should try broadly to conceptualize your research
problem before undertaking your major literature review. There are four steps
involved in conducting a literature review:
Searching for the existing literature in your area of study.
Reviewing the selected literature.
Developing a theoretical framework.
Developing a conceptual framework.
The skills required for these tasks are different. Developing theoretical and
conceptual frameworks is more difficult than the other tasks.
1. Searching for the existing literature
To search effectively for the literature in your field of enquiry, it is imperative
that you have at least some idea of the broad subject area and of the problem you
wish to investigate, in order to set parameters for your search. Next, compile a
bibliography for this broad area. There are three sou rces that you can use to
prepare a bibliography:
books; journals; the Internet.
2. Reviewing the selected literature
Now that you have identified several books and articles as useful, the next step
is to start reading them critically to pull together theme s and issues that are of
relevance to your study. Unless you have a theoretical framework of themes in
mind to start with, use separate sheets of paper for each theme or issue you
16
identify as you go through selected books and articles. While going through the
literature you should carefully and critically examine it with respect to the following
aspects:
Note whether the knowledge relevant to your theoretical framework has
been confirmed beyond doubt.
Note the theories put forward, the criticisms of these and their basis, the
methodologies adopted (study design, sample size and its characteristics,
measurement procedures, etc.) and the criticisms of them.
Examine to what extent the findings can be generalized to other situations.
Notice where there are significant differences of opinion among researchers
and give your opinion about the validity of these differences.
Ascertain the areas in which little or nothing is known - the gaps that exist
in the body of knowledge.
3. Developing a theoretical framework
Examining the literature can be a never-ending task, but as you have limited
time it is important to set parameters by reviewing the literature in relation to some
main themes pertinent to your research topic. As you start reading the literature,
you will soon discover that the problem you wish to investigate has its roots in a
number of theories that have been developed from different perspectives. The
information obtained from different books and journals now needs to be sorted
under the main themes and theories, highlighting agreements and disagreements
among the authors and identifying the unanswered questions or gaps. You will also
realise that the literature deals with a number of aspects that have a direct or
indirect bearing on your research topic. Use these aspects as a basis for developing
your theoretical framework. Your review of the literature should sort out the
information, as mentioned earlier, within this framework. Unless you review the
literature in relation to this framework, you will not be able to develop a focus in
your literature search: that is, your theoretical framework provides you with a guide
as you read. This brings us to the paradox mentioned previously: until you go
through the literature you cannot develop a theoretical framework, and until you
have developed a theoretical framework you cannot effectively review the literature.
The solution is to read some of the literature and then attempt to develop a
framework, even a loose one, within which you can organize the rest of the
literature you read. As you read more about the area, you are likely to change the
framework. However, without it, you will get bogged down in a great deal of
unnecessary reading and note-taking that may not be relevant to your study.
4. Developing a conceptual framework
The conceptual framework is the basis of your research problem. It stems from
the theoretical framework and usually focuses on the section(s) which become the
basis of your study. Whereas the theoretical framework consists of the theories or
issues in which your study is embedded, the conceptual framework describes the
aspects you selected from the theoretical framework to become the basis of your
17
enquiry. For instance, the theoretical framework includes all the theories that have
been put forward to explain the relationship between fertility and mortality.
However, out of these, you may be planning to test only one, say the fear of non -
survival. Similarly the conceptual framework is focused on indicators to measure
the success or failure of the strategies to enhance community responsiveness.
Hence the conceptual framework grows out of the theoretical framework and relates
to the specific research problem.
2.3.2 The Research Problem
Broadly speaking, any question that you want answered and any assumption
or assertion that you want to challenge or investigate can become a research
problem or a research topic for your study. However, it is important to remember
that not all questions can be transformed into research problems and some may
prove to be extremely difficult to study. According to Powers, Meenaghan and
Twoomey (1985: 38), 'Potential research questions may occur to us on a regular
basis, but the process of formulating them in a meaningful way is not at all an easy
task.' As a newcomer it might seem easy to formulate a problem but it requires
considerable knowledge of both the subject area and research methodology. Once
you examine a question more closely you will soon realize the complexity of
formulating an idea into a problem which is re searchable. 'First identifying and
then specifying a research problem might seem like research tasks that ought to be
easy and quickly accomplished. However, such is often not the case' (Yegidis &
Weinback 1991: 35). It is essential for the problem you formulate to be able to
withstand scrutiny in terms of the procedures required to be undertaken. Hence
you should spend considerable time in thinking it through.
The importance of formulating a research problem
The formulation of a research problem is the first and most important step of
the research process. It is like the identification of a destination before undertaking
a journey. In the absence of a destination, it is impossible to identify the shortest -
or indeed any - route. Similarly, in the absence of a clear research problem, a clear
and economical plan is impossible. To use another analogy, a research problem is
like the foundation of a building. The type and design of the building are dependent
upon the foundation. If the foundation is well designe d and strong you can expect
the building to be also. The research problem serves as the foundation of a research
study: if it is well formulated, you can expect a good study to follow. According to
Kerlinger: If one wants to solve a problem, one must generally know what the
problem is. It can be said that a large part of the problem lies in knowing what one
is trying to do. You must have a clear idea with regard to what it is that you want to
find out about and not what you think you must find.
A research problem may take a number of forms, from the very simple to the
very complex. The way you formulate a problem determines almost every step that
follows: the type of study design that can be used; the type of sampling strategy
that can be employed; the research instrument that can be used or developed; and
the type of analysis that can be undertaken. Suppose your broad area of interest is
18
operational purposes. "Knowing what data are available often serves to narrow the
problem itself as well as the technique that might be used." 2 . This would also help a
researcher to know if there are certain gaps in the theories, or whether the existing
theories applicable to the problem under study are inconsistent with each other, or
whether the findings of the different studies do not follow a pattern consistent with
the theoretical expectations and so on. All this will enable a researcher to take new
strides in the field for furtherance of knowledge i.e., he can move up starting from
the existing premise. Studies on related problems are useful for indicating the type
of difficulties that may be encountered in the present study as also the possible
analytical shortcomings. At times such studies may also suggest useful and even
new lines of approach to the present problem.
4. Developing the ideas through discussions:
Discussion concerning a problem often produces useful information. Various
new ideas can be developed through such an exercise. Hence, a researcher must
discuss his problem with his colleagues and others who have enough experience in
the same area or in working on similar problems. This is quite often known as an
experience survey. People with rich experience are in a position to enlighten the
researcher on different aspects of his proposed study and their advice and
comments are usually in valuable to the researcher. They help him sharpen his
focus of attention on specific aspects within the field. Discussi ons with such
persons should not only be confined to the formulation of the specific problem at
hand, but should also be concerned with the general approach to the given
problem, techniques that might be used, possible solutions, etc.
1. Rephrasing the research problem
Finally, the researcher must sit to rephrase the research problem into a
working proposition. Once the nature of the problem has been clearly understood,
the environment (within which the problem has got to be studied) has been defined,
discussions over the problem have taken place and the available literature has been
surveyed and examined, rephrasing the problem into analytical or operational
terms is not a difficult task. Through rephrasing, the researcher puts the research
problem in as specific terms as possible so that it may become operationally viable
and may help in the development of working hypotheses.*
2.3.3 Objectives
Having decided what to study, and knowing why s/he wants to study it, the
investigator can now formulate his study objectives. Objectives should be closely
related to the statement of the problem. For example, if the problem identified is
low utilization of health stations in a rural district, the general objective of the
study could be to assess the reasons for this l ow utilization. If we break down this
general objective into smaller and logically connected parts, then we get specific
objectives
General objectives: aim of the study in general terms
21
Examples
Possible research question: Are females smarter than males? This question
delineates the variables to be measured: gender and intelligence. Yet, it is unclear
how they will be evaluated: What method will be used to define and measure
intelligence?
Revised question: Do females age 18-35 score higher than adult males age 18-
35 on the WAIS-III? (The WAIS-III is a standardized intelligence test.) This research
question produces data that can be replicated. From there, the author can devise a
question that takes a stand.
In essence, the research question that guides the sciences and social sciences
should do the following three things:
1. Post a problem.
2. Shape the problem into a testable hypothesis.
3. Report the results of the tested hypothesis.
There are two types of data that can help shape research questions in the
sciences and social sciences: quantitative and qualitative data. While quantitative
data focuses on the numerical measurement and analysis between variables,
qualitative data examines the social processe s that give rise to the relationships,
interactions, and constraints of the inquiry
2.3.4 Research Gap
Once you have found a promising research area of interest, you must identify
research questions that are not only unresolved but whose exploration can
meaningfully contribute to existing theory and/or practice. In this section, you will
learn to identify such promising research questions by avoiding common beginner's
mistakes, understanding how to identify a "gap" in the scholarly literature, and
making sure it is worthwhile to try to fill the gap.
Don’t decide on a research method or focus on the data to be analyzed before
finding the research problem. Because many graduate students are relatively
inexperienced in designing and conducting original research, they can forget that
research questions must be formulated before determining the most appropriate
research method or focusing on the type of data to be collected. For example, the
researchers in Scenario 1 and Scenario 2 make the mistake of not determin ing the
research questions first. The researcher in Scenario 3, on the other hand, wisely
bases her research design on the nature of the research questions.
Scenario 1: Jane would like to conduct a content analysis for her master's
thesis project. She decides to investigate how the characteristics of online gaming
walkthroughs affect online community response. Since she is doing a content
analysis, Jane decides to collect data on text features such as reference to self, use
of jargon, use of sentence fragments, etc.
25
LESSON – 3
RESEARCH PROCESS
3.1 INTRODUCTION
The research process is the step-by-step procedure of developing one's
research. However, one can seldom progress in a step-by-step fashion as such. The
research process involves identifying, locating, assessing, analyzing, and then
developing and expressing your ideas. All of these activities will be based on
primary and secondary sources from which recommendations or plans are
formulated. The details steps in research process are explained below.
3.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Research process
Steps in research process
3.3 CONTENTS
3.3.1 Steps in research process
3.3.1 Steps in research process
Research process consists of a series of steps or actions required for effectively
conducting research. The following are the steps that provide useful procedural
guidelines regarding the conduct of research:
1. Formulating the research problem;
2. Extensive literature survey;
3. Developing hypothesis;
4. Preparing the research design;
5. Determining sample design;
6. Collecting data;
7. Execution of the project;
8. Analysis of data;
9. Hypothesis testing;
10. Generalization and interpretation, and
11. Preparation of the report or presentation of the results.
1. Formulating the research problem
The first and foremost stage in the research process is to select and properly
define the research problem. A researcher should first identify a problem and
formulate it, so as to make it amenable or susceptible to research. In general, a
research problem refers to an unanswered question that a researcher might
encounter in the context of either a theoretical or practical situation, which he/she
would like to answer or find a solution to. Identification of a research problem is the
pre-condition to conducting research. A research problem is said to be the one
30
which requires a researcher to find the best available solution to the given problem.
That is, the researcher needs to find out the best course of action through which
the research objective may be achieved optimally in the context of a given situation.
Several factors may contribute to making the problem complicated. For example,
the environment may alter, thus affecting the efficiencies of the alternative courses
of action taken or the quality of the outcomes. The number of alternative courses of
action might be very large and the individual not involved in making the decision
may be affected by the change in environment and may react to it favorably or
unfavorably. Other similar factors are also likely to cause such changes in the
context of research, all of which may be considered from the point of view of a
research problem.
2. Extensive literature survey
Once the problem is formulated, a brief summary of it should be written down.
It is compulsory for a research worker writing a thesis for a Ph.D. degree to write a
synopsis of the topic and submit it to the necessary Committee or the Research
Board for approval. At this juncture the researcher should undertake extensive
literature survey connected with the problem. For this purpose, the abstracting and
indexing journals and published or unpublished bibliographies are the first place to
go to. Academic journals, conference proceedings, government reports, books etc.,
must be tapped depending on the nature of the problem. In this process, it should
be remembered that one source will lead to another. The earlier studies, if any,
which are similar to the study in hand should be carefully studied. A good library
will be a great help to the researcher at this stage.
3. Development of working hypotheses
After extensive literature survey, researcher should state in clear terms the
working hypothesis or hypotheses. Working hypothesis is tentative assumption
made in order to draw out and test its logical or empiric al consequences. As such
the manner in which research hypotheses are developed is particularly important
since they provide the focal point for research. They also affect the manner in which
tests must be conducted in the analysis of data and indirectly the quality of data
which is required for the analysis. In most types of research, the development of
working hypothesis plays an important role. Hypothesis should be very specific and
limited to the piece of research in hand because it has to be tested. The role of the
hypothesis is to guide the researcher by delimiting the area of research and to keep
him on the right track. It sharpens his thinking and focuses attention on the more
important facets of the problem. It also indicates the type of data require d and the
type of methods of data analysis to be used.
How does one go about developing working hypotheses? The answer by using
the following approach:
a) Discussions with colleagues and experts about the problem, its origin and
the objectives in seeking a solution;
31
or the data, with the help of which he examines the truth con tained in his
hypothesis. But in the case of a survey, data can be collected by any one or more of
the following ways:
By observation, through personal interview, through telephone interview,
mailing of questionnaire, through schedules:
The researcher should select one of these methods of collecting the data taking
into consideration the nature of investigation, objective and scope of the inquiry,
financial resources, available time and the desired degree of accuracy. Though he
should pay attention to all these factors but much depends upon the ability and
experience of the researcher. In this context Dr A.L. Bowley very aptly remarks that
in collection of statistical data commonsense is the chief requisite and experience
the chief teacher.
7. Execution of the project
Execution of the project is a very important step in the research process. If the
execution of the project proceeds on correct lines, the data to be collected would be
adequate and dependable. The researcher should see that the project is exec uted in
a systematic manner and in time. If the survey is to be conducted by means of
structured questionnaires, data can be readily machine -processed. In such a
situation, questions as well as the possible answers may be coded. If the data are to
be collected through interviewers, arrangements should be made for proper
selection and training of the interviewers. The training may be given with the help
of instruction manuals which explain clearly the job of the interviewers at each
step. Occasional field checks should be made to ensure that the interviewers are
doing their assigned job sincerely and efficiently. A careful watch should be kept for
unanticipated factors in order to keep the survey as much realistic as possible.
This, in other words, means that steps should be taken to ensure that the survey is
under statistical control so that the collected information is in accordance with the
pre-defined standard of accuracy. If some of the respondents do not cooperate,
some suitable methods should be designe d to tackle this problem. One method of
dealing with the non-response problem is to make a list of the non -respondents and
take a small sub-sample of them, and then with the help of experts vigorous efforts
can be made for securing response.
8. Analysis of data
After the data have been collected, the researcher turns to the task of
analyzing them. The analysis of data requires a number of closely related
operations such as establishment of categories, the application of these categories
to raw data through coding, tabulation and then drawing statistical inferences. The
unwieldy data should necessarily be condensed into a few manageable groups and
tables for further analysis. Thus, researcher should classify the raw data into some
purposeful and usable categories. Coding operation is usually done at this stage
through which the categories of data are transformed into symbols that may be
tabulated and counted. Editing is the procedure that improves the quality of the
34
data for coding. With coding the stage is re ady for tabulation. Tabulation is a part
of the technical procedure wherein the classified data are put in the form of tables.
The mechanical devices can be made use of at this juncture. A great deal of data,
specially in large inquiries, is tabulated by c omputers. Computers not only save
time but also make it possible to study large number of variables affecting a
problem simultaneously.
Analysis work after tabulation is generally based on the computation of
various percentages, coefficients, etc., by appl ying various well defined statistical
formulae. In the process of analysis, relationships or differences supporting or
conflicting with original or new hypotheses should be subjected to tests of
significance to determine with what validity data can be said to indicate any
conclusion(s). For instance, if there are two samples of weekly wages, each sample
being drawn from factories in different parts of the same city, giving two different
mean values, then our problem may be whether the two mean values are
significantly different or the difference is just a matter of chance. Through the use
of statistical tests we can establish whether such a difference is a real one or is the
result of random fluctuations. If the difference happens to be real, the inference w ill
be that the two samples come from different universes and if the difference is due to
chance, the conclusion would be that the two samples belong to the same universe.
Similarly, the technique of analysis of variance can help us in analysing whether
three or more varieties of seeds grown on certain fields yield significantly different
results or not. In brief, the researcher can analyse the collected data with the help
of various statistical measures.
9. Hypothesis testing
After analyzing the data as stated above, the researcher is in a position to test
the hypotheses, if any, he had formulated earlier. Do the facts support the
hypotheses or they happen to be contrary? This is the usual question which should
be answered while testing hypotheses. Various tests, such as Chi square test, t-
test, F-test, have been developed by statisticians for the purpose. The hypotheses
may be tested through the use of one or more of such tests, depending upon the
nature and object of research inquiry. Hypothesis-testing will result in either
accepting the hypothesis or in rejecting it. If the researcher had no hypotheses to
start with, generalizations established on the basis of data may be stated as
hypotheses to be tested by subsequent researches in times to come.
10. Generalizations and interpretation
If a hypothesis is tested and upheld several times, it may be possible for the
researcher to arrive at generalization, i.e., to build a theory. As a matter of fact, the
real value of research lies in its ability to arrive at certain generalizations. If the
researcher had no hypothesis to start with, he might seek to explain his findings on
the basis of some theory. It is known as interpretation. The process of
interpretation may quite often trigger off new questions which i n turn may lead to
further researches.
35
2. http://as.wiley.com/Wile yCDA/WileyTitle/productCd-1118682092.html
3. http://scholarcommons.usf.edu/cgi/viewcontent.cgi?article=1002&c ontext
=oa_textbooks
3.9 ASSIGNMENTS
1. “Empirical research in India in particular creates so many problems for the
researchers”. State the problems that are usually faced by such researchers
3.10 SUGGESTED READINGS/ REFERENCE BOOKS
1. Balnaves, M. and Caputi, P. (2001) Introduction to Quantitative Research
Methods: An Investigative Approach, London: Sage.
2. Bernard, H.R. (2000) Social Research Methods: Qualitative and Quantitative
Approaches, Thousand Oaks, CA: Sage.
3. Kothari, C.R.,1985, Research Methodology- Methods and Techniques, New
Delhi, Wiley Eastern Limited.
Kumar, Ranjit, 2005, Research Methodology-A Step-by-Step Guide for
Beginners,(2nd.ed.),Singapore, Pearson Education.
3.11 LEARNING ACTIVITIES
1. Plan an descriptive research project and describe the various steps which
are followed in completing the project
3.12 KEYWORDS
Bias is a deliberate attempt either to conceal or highlight something that you
found in your research or to use deliberately a procedure or method that you know
is not appropriate but will provide information that you are looking for because you
have a vested interest in it.
Control group: The group in an experimental study which is not exposed to
the experimental intervention is called a control group. The sole purpose of the
control group is to measure the impact of extraneous and chance variables on the
dependent variable.
Editing consists of scrutinising the completed research instruments to identify
and minimise, as far as possible, errors, incompleteness, misclassification and gaps
in the information obtained from respondents.
Pie chart: The pie chart is another way of representing data graphically. As
there are 360 degrees in a circle, the full circle can be used to represent 100 per
cent or the total population. The circle or pie is divided into sections in accordance
with the magnitude of each subcategory comprising the total population. Hence
each slice of the pie is in proportion to the size of each subcategory of a frequency
distribution.
37
LESSON – 4
RESEARCH DESIGN
4.1 INTRODUCTION
The most important problem after defining the research problem is preparing
the design of the research project, which is popularly known as the 'research
design'. A research design helps to decide upon issues like what, when, where, how
much, by what means, etc., with regard to an enquiry or a research study. "A
research design is the arrangement of conditions for collection and analysis of data
in a manner that aims to combine relevance to the research purpose with economy
in procedure. In fact, the research design is the conceptual structures within which
research is conducted; it constitutes the blueprint for the collection, measurement
and analysis of data" (Selltiz, et.al. 1962). Thus, research design provi des an outline
of what the researcher is going to do in terms of framing the hypothesis, its
operational implications, and the final data analysis. Specifically, the research
design highlights decisions which include:
i) The nature of the study
ii) The purpose of the study
iii) The location where the study would be conducted
iv) The nature of data required
v) From where the required data can be collected
vi) What time period the study would cover
vii) The type of sample design that would be used
viii) The techniques of data collection that would be used
ix) The methods of data analysis that would be adopted
x) The manner in which the report would be prepared
In view of the stated research design decisions, the overall research design
may be divided into the following (Kothari 1988)
a) The sampling design that deals with the method of selecting items to be
observed for the selected study;
b) The observational design that relates to the conditions under which the
observations are to be made;
c) The statistical design that concerns with the question of how many items
are to be observed, and how the information and data gathered are to be
analysed; and
d) The operational design that deals with the techniques by which the
procedures specified in the sampling, statistical and observational designs
can be carried out.
38
4.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Research design
Features of research design
Characteristics of good research design
Types of research design
4.3 CONTENTS
4.3.1 Features of research design
4.3.2 Characteristics of research design
4.3.3 Importance of research design
4.3.4 Types
4.3.1 Features of research design
The important features of research design may be outlined as follows:
i) It constitutes a plan that identifies the types and sources of information
required for the research problem;
ii) It constitutes a strategy that specifies the methods of data collection and
analysis which would be adopted; and
iii) It also specifies the time period of research and monetary budget involved
in conducting the study, which comprise the two major constraints of
undertaking any research.
4.3.2 Characteristics of a good research design
A good research design often possesses the qualities such as being flexible,
suitable, efficient, economical, and so on. Generally, a research design which
minimizes bias and maximizes the reliability of the data collected and analyzed is
considered a good design (Kothari 1988).
A research design which involves the smallest experimental error is said to be
the best design for investigation. Further, a researc h design that yields maximum
information and provides an opportunity of viewing the various dimensions of a
research problem is considered to be the most appropriate and efficient design.
Thus, the question of a good design relates to the purpose or objective and nature
of the research problem studied. While a research design may be good, it may not
be equally suitable to all studies. In other words, it may be lacking in one aspect or
the other in the case of some other research problems. Therefore, no sing le research
design can be applied to all types of research problems.
A research design suitable for a specific research problem would usually
involve the following considerations:
i) The methods of gathering the information;
ii) The skills and availability of the researcher and his/her staff, if any;
iii) The objectives of the research problem being studied;
39
LESSON – 5
SOURCES OF DATA
5.1 INTRODUCTION – MEANING, TYPES AND IMPORTANCE
“Data are facts, figures and other relevant materials past and present serving
as bases for study and analysis”.
Meaning of Data
The search for answers to research questions calls collection of Data. “Data are
facts, figures and other relevant materials, past and present, serving as bases for
study and analysis”.
Types of Data
The Data needed for social science Research may be broadly classified into:
a) Data pertaining to human beings
b) Data relating to Organisations
c) Data pertaining to territorial area.
A) Personal Data (relating to Human beings) are of two types.
1. Demographic and socio-economic characteristics of individuals. Like name,
sex, race, social class, relation, education, occupation, income etc.
2. Behavioural Variables: Attitudes, opinion knowledge, practice, intensions
etc.
B) Organistion Data:- Consist of data relating to an organizations, origin
ownership, function, performance etc.
C) Territorial Data: - are related to geo-physical characteristic, population,
infrastructure etc of divisions like villages, cities, taluks, distinct, state etc.
Importance of Data
The data serve as the bases or raw materials for analysis without Data no
specific inferences can be drawn on our study. The reliability of data determines the
quality of research. Data form the basis of testing hypothesis data provides the
facts and figures for constructing measuring scale. The scientific process of
research can be carried out only through accurate data.
5.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Sources of data
Types of data
Various data collection methods
5.3 CONTENTS
5.3.1 Sources of Data
5.3.2 Data Collection Methods
44
Panels
Panels, like focus groups, are another source of primary information for
research purposes. Whereas focus groups meet for a one -time group session,
panels (of members) meet more than once. In cases where the effects of certain
interventions or changes are to be studied over a period of time, panel studies are
very useful. Individuals are randomly chosen to serve as panel members for a
research study. For instance, if the effects of a proposed advertisement for a certain
brand of coffee are to be assessed quickly, the panel members can be exposed to
the advertisement and their intentions of purchasing that brand assessed. This can
be taken as the response that could be expected of consumers if, in fact, they had
been exposed to the advertisement. A few months later, the product manager might
think of introducing a change in the flavor of the same product and explore its
effects on this panel. Thus, a continuing set of "experts" serves as the sample base
or the sounding board for assessing the effects of change. Such expert members
compose the panel, and research that uses them is called a panel study.
Unobtrusive Measures
Trace measures, or unobtrusive measures as they are also called, originate
from a primary source that does not involve people. One example is the wear and
tear of journals in a university library, which offers a good indication of their
popularity, frequency of use, or both. The number of di fferent brands of soft drink
cans found in trash bags also provides a measure of their consumption levels.
Signatures on checks exposed to ultraviolet rays could indicate the extent of forgery
and frauds; actuarial records are good sources for collecting d ata on the births,
marriages, and deaths in a community; company records disclose a lot of personal
information about employees, the level of company efficiency, and other data as
well. Thus these unobtrusive sources of data and their use are also importan t in
research.
Secondary Sources
Secondary data are indispensable for most organizational research. Secondary
data refer to information gathered by someone other than the researcher
conducting the current study. Such data can be internal or external to the
organization and accessed through the Internet or perusal of recorded or published
information.
Secondary data can be used, among other things, for forecasting sales by con -
structing models based on past sales figures, and through extrapolation. There are
several sources of secondary data, including books and periodicals, government
publications of economic indicators, census data, Statistical Abstracts, data bases,
the media, annual reports of companies, etc. Case studies, and other archival
records—sources of secondary data—provide a lot of information for research and
problem solving. Such data are, as we have seen, mostly qualitative in nature. Also
included in secondary sources are schedules maintained for or by key personnel in
organizations, the desk calendar of executives, and speeches delivered by them.
Much of such internal data, though, could be proprietary and not accessible to all.
46
Financial databases readily available for research are also secondary data
sources. The Compustat Database contains information on thousands of companies
organized by industry, and information on global companies is also avail able
through Compustat.
The advantage of seeking secondary data sources is savings in time and costs
of acquiring information. However, secondary data as the sole source of information
has the drawback of becoming obsolete, and not meeting the specific needs of the
particular situation or setting. Hence, it is important to refer to sources that offer
current and up-to-date information.
5.3.2 Data collection methods
Having examined the various sources of data, let us now look into the data
collection methods.
The important data collection methods are:
1. Observation
2. Interviewing
3. Questionnaire
4. Schedules
1. Observational method
Whereas interviews and questionnaires elicit responses from the subjects, it is
possible to gather data without asking questions of respondents. People can be
observed in their natural work environment or in the lab setting, and their activities
and behaviors or other items of interest can be noted and recorded.
Apart from the activities performed by the individuals under study, their move -
ments, work habits, the statements made and meetings conducted by them, their
facial expressions of joy, anger, and other emotions, and body language can be
observed. Other environmental factors such as layout, work-flow patterns, the
closeness of the seating arrangement, and the like, can also be noted. Children can
be observed as to their interests and attention span with various stimuli, such as
their involvement with different toys. Such observation would help toy manufac -
turers, child educators, day-care administrators, and others deeply involved in or
responsible for children's development, to design and model ideas based on c hil-
dren's interests, which are more easily observed than traced in any other manner.
The researcher can play one of two roles while gathering field observational data—
that of a nonparticipant-observer or participant-observer.
Advantages and Disadvantages of Observational Studies
There are some specific advantages and disadvantages to gathering data
through observation as listed below.
Advantages of Observational Studies
The following are among the advantages of observational studies.
47
1. The data obtained through observation of events as they normally occur are
generally more reliable and free from respondent bias.
2. In observational studies, it is easier to note the effects of environmental
influences on specific outcomes. For example, the weather (hot, c old, rainy), the day
of the week (midweek as opposed to Monday or Friday), and such other factors that
might have a bearing on, for example, the sales of a product, traffic patterns,
absenteeism, and the like, can be noted and meaningful patterns might emerge
from this type of data.
3. It is easier to observe certain groups of individuals—for example, very young
children and extremely busy executives—from whom it may be otherwise difficult to
obtain information.
Drawbacks of Observational Studies
The following drawbacks of observational studies have also to be noted.
1. It is necessary for the observer to be physically present (unless a camera or
another mechanical system can capture the events of interest), often for prolonged
periods of time.
2. This method of collecting data is not only slow, but also tedious and
expensive.
3. Because of the long periods for which subjects are observed, observer
fatigue could easily set in, which might bias the recorded data.
4. Though moods, feelings, and attitudes can be guessed by observing facial
expressions and other nonverbal behaviors, the cognitive thought processes of
individuals cannot be captured.
5. Observers have to be trained in what and how to observe, and ways to avoid
observer bias.
2. Interview
One method of collecting data is to interview respondents to obtain
information on the issues of interest. Interviews could be unstructured or
structured, and conducted either face to face or by telephone or online. The
unstructured and structured interviews are discussed first. Some important factors
to be borne in mind while interviewing will then be detailed; the advan tages and
disadvantages of face-to-face interviewing and telephone interviews enumerated
thereafter; and finally, computer-assisted interviews described.
Unstructured and Structured Interviews
Unstructured Interviews
Unstructured interviews are so labeled because the interviewer does not enter
the interview setting with a planned sequence of questions to be asked of the
respondent. The objective of the unstructured interview is to bring some pre -
liminary issues to the surface so that the researcher can determine what variables
need further in-depth investigation. In the discussion of the "Broad Problem Area,"
we saw several situations where the man ager might entertain a vague idea of
48
certain changes taking place in the situation without knowing what exactly they
are. Such situations call for unstructured interviews with the people concerned. In
order to understand the situation in its totality, the researcher will interview
employees at several levels. In the initial stages, only broad, open -ended questions
would be asked, and the replies to them would inform the researcher of the
perceptions of the individuals. The type and nature of the questions asked of the
individuals might vary according to the job level and type of work done by them.
For instance, top and middle-level managers might be asked more direct questions
about their perceptions of the problem and the situation. Employees at lower leve ls
may have to be approached differently.
Clerical and other employees at lower hierarchical levels may be asked broad,
open-ended questions about their jobs and the work environment during the
unstructured interviews. Supervisors may be asked broad questi ons relating to
their department, the employees under their supervision, and the organization.
Structured Interviews
Structured interviews are those conducted when it is known at the outset what
information is needed. The interviewer has a list of predetermined questions to be
asked of the respondents either personally, through the telephone, or through the
medium of a PC. The questions are likely to focus on factors that had surfaced
during the unstructured interviews and are considered relevant to the problem. As
the respondents express their views, the researcher would note them down. The
same questions will be asked of everybody in the same manner. Sometimes,
however, based on the exigencies of the situation, the experienced researcher might
take a lead from a respondent's answer and ask other relevant questions not on the
interview protocol. Through this process, new factors might be identified, resulting
in a deeper understanding. However, to be able to recognize a probable response,
the interviewer must comprehend the purpose and goal of each question. This is
particularly important when a team of trained interviewers conducts the survey.
Face-to-Face and Telephone Interviews
Interviews can be conducted either face to face or over the telephone. They
could also be computer-assisted. Although most unstructured interviews in orga-
nizational research are conducted face to face, structured interviews could be either
face to face or through the medium of the telephone, depending on the level of
complexity of the issues involved, the likely duration of the interview, the
convenience of both parties, and the geographical area covered by the survey.
Telephone interviews are best suited when information from a large number of
respondents spread over a wide geographic area is to be obtained quickly, and the
likely duration of each interview is, say, 10 minutes or less. Many market surveys,
for instance, are conducted through structured telephone interviews. In addition,
computer-assisted telephone interviews (CATI) are also possible, and easy to
manage.
49
the respondent is asked to write his answer. This type of item permits explanation,
but responses can be difficult to summarize and tabulate. The responses also may
be too brief, or the respondent may have omitted important information.
The open form item is of most value when unrestricted, in depth responses are
needed. An example of this type of questionnaire item follows:
In your opinion, what might your high school have done to prepare you more
adequately for college?
The closed form item sometimes is referred to as the "restricted" or
"structured" type. It consists of a question or a statement to which a person
responds by selecting one or more choices, such as "Yes" or "No". In one variation of
this type the respondent may be asked to underline award from the two or more
alternatives. Another variation requires the ranking of choices.
The close form item facilities the tabulation and analysis of data. It also
improves the reliability and consistency of the data. One limitation of this kind of
item is that the respondent does not have the opportunity to explain why he has
given certain responses, and this may be important in some kinds of research
studies. The close form item also limits the scope and depth of responses, so its use
in measuring attitudes, feelings and certain aspects of behavior may be limited. It is
possible also that the answers from which the respondents must choose are not the
proper ones for all of the individuals who are given the questionnaire. An example
of a closed form questionnaire item follows: If group tests are used in your school,
by whom are they administered? (a) Administrators (b) Counselors,
(c) Psychologists, (d) Psychometricians,
(e) Teachers, and (e) Others
Elements of Questioning Method
In using the questioning method there are six ele ments to be considered:
i) The content of the questioning; which includes both why the researcher is
asking the questions and why the respondent believes they are being asked,
since these are not necessarily the same;
ii) Content of the questioning;
iii) The question which can be defined as the verbal stimulus to which the
respondent will be exposed;
iv) The provision for answering which refers to the response the respondent will
be.
v) The way in which the response will be recorded, which refers to variation like
writing or speaking or recording;
vi) The nature of the researcher respondent interaction, which considers
whether researcher respondent interaction , which considers whether
researcher and respondent actually meets face to face or whether they
communicate impersonally.
53
that students constructing questionnaires will become critic al of their first efforts
and strive to make each question as clear as possible.
1. Define or qualify terms that could easily be a misinterpreted. "What is the
value of your horse? The meaning of the term value is not clear. These
values may differ considerably. It is essential to frame specific questions
such as, "what is the present market value of your horse?"
2. Be careful in using descriptive adjectives and adverbs that have no agreed -
upon-meaning. This fault is frequently found in rating scales as well as
questionnaire.
3. Beware of double negatives (1) Federal aid should not granted for those
states in which education' is not equal regardless of race, creed, or colour.
4. Be careful of inadequate alternatives (1) Married? Yes/No.
5. Avoid the double barreled question. Example: Do you believe that gifted
students should be placed in separate groups for instructional purposes
and assigned to special schools?
6. Underline the word if you wish to indicate special emphasis. Example
7. Should all schools offer modern foreign language?
8. When asking for ratings or comparisons a point of reference is necessary.
Example: How would you rate this student teacher's class room teaching?
Superior Average-Below Average .
Phrase questions so that they are appropriate for all respondents. Desi gn
questions that will give a complete response. Such questions which as answer "Yes
or no" would not reveal much information about the reading habits of the
respondents.
There must be a provision for the systematic quantification of response. One
type of question that asks respondents to check a number of items from a list is
difficult to summarize especially all respondents do not check the same number.
The students should bear in mind that these suggestions are used in constructing
questionnaire items.
Scope of Questionnaire
In descriptive studies where the sources are varied and widely scattered, the
questionnaire is a major instrument for gathering data. It is very handy in cases,
where one cannot conveniently see personally all the people from whom the
responses are required. This technique finds favour in determination of present
status in certain aspects of education-current practice in schools, financial aspects
service conditions of teachers, etc. It can be used over extensive range of territory -
national and international. "As research techniques are becoming more and more
refined day after day, it is hoped that this wayward child of the science of education
will soon curb its unruly disposition and also mend its unseemly ways".
57
12. A good questionnaire will arouse the curiosity as regards resu lts and will
stimulate them to make supplementary inquiries and promise them a copy
of the published results.
13. It should be so organized as to avoid overlapping of items in different
sections.
14. Evidently respondents will not like to incur expenditure on postage for
returning the questionnaire. The investigator should therefore see that it
accompanied by self-addressed stamped envelope, in case; it is to be
returned by post.
Limitations of the Questionnaire Technique
In this technique, a research worker has to depend on several hundred
persons from whom response is expected and it is not an easy job to get active and
willing cooperation of all the respondents. One may be very diligent and sincere
about his work but one cannot be sure that the responses would be forthcoming.
Some of the respondents may hold back their replies because they are skeptical
about the value of research, others may not respond for want of time, or because
they do not feel interested in the problem in hand or because they have not been
sufficiently motivated by the introductory letter. The research worker may therefore
remain in a state of expectancy and his work may be delayed.
1. Even when he gets back a sufficiently large number of questionnaires, he
may find that all of them have not been completely filled in Omissions in
some cases may be such as will materially vitiate the results and
significantly affect the interpretations.
2. If the questionnaire is sent to different areas or to people of different
categories and one set of respondents deliberately with holds its responses,
the inquiry would be affected adversely. Suppose there is a judicial inquiry
on discrimination against a minority and section of the minority concerned
refuses to cooperate as a protest, evidently the results of the i nquiry will
not present a true picture.
3. Sometimes the respondents fill in their responses very indifferently,
without bothering about their correctness and sometimes they deliberately
give wrong information.
4. Interview Schedule
Scheduling is defined as the translation of the developed plan a time table,
showing the calendar date for the start and completion of work. The scheduled start
and end of each activity or work package as well as the total project is emphasized.
The schedule helps to determine the operating is emphasized. The schedule helps
to determine the operating budget for the project and permits us to allocate
resources to' the activities.
It is within the scheduling process that we become concerned with competition
for one or more of the resources that may exist in an organization. In establishing a
60
schedule for the project, we are concerned not only with the time to do a job but
also with the exact data that the office involved may be able to do our work, having
considered all the other projects in the organization that will utilize this same
resources. The planned schedule, which is generated as an output of the
scheduling process, enable the project director to judge event progress and forecast
a data of completion. The scheduling' process also enables the project director to
inform various organizational units of the schedules that they must maintain for
their particular task.
Some Scheduling Constraints
On the surface, the process of scheduling appears relatively simple. Several
constraints, however, make the process somewhat difficult and often lead to the
development of less than an ideal or optimum schedule. Some constraints are given
in the following list:
1. The availability of particular resources during specific calendar periods.
2. The general sequence of the work in the project plan.
3. Consideration of resource requirements of other present or future projects.
4. Different or conflicting demands on the same resource.
5. A desire to avoid peak load for particular skills.
6. The available local capacity to do a particular task.
7. Limitations and requirements imposed by funding agencies.
8. Desire to minimize overtime and idle time.
9. Necessary integration with other plans or projects using the same
resources.
10. The manager's judgement of a reasonable time for pe rforming activities of
an uncertain nature.
11. Technical constraints such as uncertainties which may require extra time.
12. Local personnel policies concerning work practices (vacations, sick leave,
etc.)
13. National, states, and local laws governing work practices .
14. Difficulties inherent in scheduling far in advance.
15. The varying number of working days in a month and their translation into
calendar dates.
16. Many of these constraints relate to project scheduling in the business-
military complex, but have some application to the educational situation.
The educational situation, in contrast, has some unique constraints which
can affect schedules. Notable among these is the nine to ten month period
in which educational personnel operate.
61
under different project managers. These projects draw upon several common
resources of the organization. Consequently conflicts arise about the utilization of
the resources. The employment of the previous procedures makes the work flow for
a particular department more even and insures that the work of the projects will be
accomplished according to the schedule.
Limitations of a Schedule
The following are the main disadvantages of a schedule:
1. It is very time consuming and costly instrument in admini stering to the
subject personally.
2. Sometimes some subjects have several time queries about the schedule
and difficult to explain and satisfy them.
3. Some subjects are to be conducted to get-data.
4. Some of the subjects e.g. principals administrators are not easily
approachable and get appointment for administering tool.
5. Sometimes subjects are more alert or intelligent than the researcher.
Researcher has the difficulty to administer the tools.
6. On a large sample of subjects this tool can not be used effectively and
easily.
5.4 REVISION POINTS
1. Data are facts, figures and other relevant materials, past and present,
serving as bases for study and analysis
2. Primary sources are original sources from which the researcher directly
collects data that have not been previously collected.
3. The secondary sources consist of readily available and compiled statistical
statements and reports
4. The focus sessions are aimed at obtaining respondents' impressions,
interpretations, and opinions, as the members talk about the event,
concept, product, or service
5. Panels, like focus groups, are another source of primary information for
research purposes
6. Secondary data are indispensable for most organizational research
5.5 INTEXT QUESTIONS
1. Explain primary and secondary data and distinguish between them.
2. Explain the different methods of collecting primary data.
3. Explain direct personal, indirect oral interview, information received
4. through agencies etc.)
5. Explain the merits and demerits of different methods of collecting primary
data.
64
LESSON – 6
status. In marketing research, ordinal scales are used to measure relative attitudes,
opinions, perceptions and preferences. Measurements of this type include ‘greater
than’ or ‘less than’ judgments from the respondents. In an ordinal scale, as in a
nominal scale, equivalent objects receive the same rank. Any series of numbers can
be assigned that preserves the ordered relationships between the objects’ In other
words; any monotonic positive (order preserving) transformation of the scale is
permissible, since the differences in numbers are void of any meaning other than
order (see the following example). For these reasons, in addition to the counting
operation allowable for nominal scale data, ordinal scales permit the use of
statistics based on percentiles. It is me aningful to calculate percentile, quartile,
median, rank-order correlation or other summary statistics from ordinal data.
Interval scale
In an interval scale, numerically equal distances on the scale represent equal
values in the characteristic being measured. An interval scale contains all the
information of an ordinal scale, but it also allows you to compare the differences
between objects. The difference between any two scale values is identical to the
difference between any other two adjacent values of an interval scale. There is a
constant or equal interval between scale values. The difference between 1 and 2 is
the same as the difference between 2 and 3, which is the same as the difference
between 5 and 6. A common example in everyday life is a temperature scale. In
marketing research, attitudinal data obtained from rating scales are often treated
as interval data. In an interval scale, the location of the zero point is not fixed.
Both the zero point and the units of measurement are arbitrary. Hence, any positive
linear transformation of the form y = a + bx will preserve the properties of the scale.
Here, x is the original scale value, y is the transformed scale value, b is a positive
constant, and a is any constant. Therefore, two interval scales that rate objects A,
B, C and D as 1, 2, 3 and 4 or as 22, 24, 26 and 28 are equivalent. Note that the
latter scale can be derived from the former by using a = 20 and b = 2 in the
transforming equation. Because the zero point is not fixed, it is not meaningful to
take ratios of scale values. As can be seen, the ratio of D to B values changes from
2:1 to 7:6 when the scale is transformed. Yet, ratios of differences between scale
values are permissible. In this process, the constants a and b in the transforming
equation drop out in the computations. The ratio of the difference between D and B
values to the difference between C and B values is 2:1 in both the scales. Statistical
techniques that may be used on interval scale data include all those that can be
applied to nominal and ordinal data in addition to the arithmetic mean, standard
deviation (Chapter 18), product moment correlations, and other statistics
commonly used in marketing research. Certain specialized statistics such as
geometric mean, harmonic mean and coefficient of variation, however, are not
meaningful on interval scale data.
Ratio scale
A ratio scale possesses all the properties of the nominal, ordinal and interval
scales, and, in addition, an absolute zero point. Thus, in ratio scales we can
70
identify or classify objects, rank the objects, and compare intervals or differences. It
is also meaningful to compute ratios of scale values. Not only is the difference
between 2 and 5 the same as the difference between 14 and 17, but also 14 is
seven times as large as 2 in an absolute sense. Common examples of ratio scales
include height, weight, age and money. In marketing, sales, costs, market share
and number of customers are variables measured on a ratio scale. Ratio scales
allow only proportionate transformations of the form y = bx, where b is a positive
constant. One cannot add an arbitrary constant, as in the case of an interval scale.
An example of this transformation is provided by the conversion of yards to feet (b =
3). The comparisons between the objects are identical whether made in yards or
feet. All statistical techniques can be applied to ratio data. These include specialized
statistics such as geometric mean, harmonic mean and coefficient of variation.
6.3.2 Measurement accuracy
A measurement is a number that reflects some characteristic of an object. A
measurement is not the true value of the characteristic of interest but rather an
observation of it. A variety of factors can cause measurement error, which results in
the measurement or observed score being different from the true score of the
characteristic being measured. The true score model provides a framework for
understanding the accuracy of measurement.
According to this model, XO = XT + XS + XR
where
XO = The Observed Score Or Measurement
XT = The True Score Of The Characteristic
XS = Systematic Error
XR = Random Error
Note that the total measurement error includes the systematic error, XS , and
the random error, XR . Systematic error affects the measurement in a constant
way. It represents stable factors that affect the observed score in the same way each
time the measurement is made, such as mechanical factors. Random error, on the
other hand, is not constant. It represents transient factors that affect the observed
score in different ways each time the measurement is made, such as short-term
transient personal factors or situational factors . The distinction between
systematic and random error is crucial to our understanding of reliability and
validity.
Reliability It refers to the extent to which a scale produces consistent results
if repeated measurements are made. Systematic sources of error do not have an
adverse impact on reliability, because they affect the measurement in a constant
way and do not lead to inconsistency. In contrast, random error produces
inconsistency, leading to lower reliability. Reliability can be defined as the extent to
which measures are free from random error, XR . If XR = 0, the measure is perfectly
reliable. Reliability is assessed by determining the proportion of systematic
71
this type, each item measures some aspect of the construct measured by the entire
scale, and the items should be consistent in what they indicate about the
construct. This measure of reliability focuses on the internal consistency of the set
of items forming the scale. The simplest measure of internal consistency is split-
half reliability. The items on the scale are divided into two halves and the resulting
half scores are correlated. High correlations between the variables indicate high
internal consistency. The scale items can be split into halves based on odd - and
even-numbered items or randomly. The problem is that the results will depend on
how the scale items are split. A popular approach to overcoming this problem is to
use the coefficient alpha. The coefficient alpha, or Cronbach’s alpha, is the average
of all possible split-half coefficients resulting from different ways of splitting the
scale items. This coefficient varies from 0 to 1, and a value of 0.6 or less generally
indicates unsatisfactory internal consistency reliability. An important property of
coefficient alpha is that its value tends to i ncrease with an increase in the number
of scale items. Therefore, coefficient alpha may be artificially, and inappropriately,
inflated by including several redundant scale items.
Another coefficient that can be employed in conjunction with coefficient alpha
is coefficient beta. Coefficient beta assists in determining whether the averaging
process used in calculating coefficient alpha is masking any inconsistent items.
Some multi-item scales include several sets of items designed to measure different
aspects of a multidimensional construct. For example, bank image is a
multidimensional construct that includes country of origin, range of products,
quality of products, service of bank personnel, credit terms, investment rates,
convenience of location, and physical layout of branches. Hence, a scale designed to
measure bank image would contain items measuring each of these dimensions.
Because these dimensions are somewhat independent, a measure of internal
consistency computed across dimensions would be inappropriate. If several items
are used to measure each dimension, however, internal consistency reliability can
be computed for each dimension.
Validity
The validity of a scale may be considered as the extent to which differences in
observed scale scores reflect true differences among objects on the characteristic
being measured, rather than systematic or random error. Perfect validity requires
that there be no measurement error (XO = XT , XR = 0, XS = 0). Researchers may
assess content validity, criterion vali dity or construct validity.
Content validity, sometimes called face validity, is a subjective but systematic
evaluation of how well the content of a scale represents the measurement task at
hand. The researcher or someone else examines whether the scale items adequately
cover the entire domain of the construct being measured. Thus, a scale designed to
measure bank image would be considered inadequate if it omitted any of the major
dimensions (range of products, quality of products, service of bank personnel , etc.).
Given its subjective nature, content validity alone is not a sufficient measure of the
73
validity of a scale, yet it aids in a common -sense interpretation of the scale scores.
A more formal evaluation can be obtained by examining criterion validity
Criterion validity reflects whether a scale performs as expected in relation to
other selected variables (criterion variables) as meaningful criteria. If, for example, a
scale is designed to measure loyalty in customers, criterion validity might be
determined by comparing the results generated by this scale with results generated
by observing the extent of repeat purchasing. Based on the time period involved,
criterion validity can take two forms, concurrent validity and predictive validity.
Concurrent validity is assessed when the data on the scale being evaluated
(e.g. loyalty scale) and the criterion variables (e.g. repeat purchasing) are collected
at the same time. The scale being developed and the alternative means of
encapsulating the criterion variables would be administered simultaneously and
the results compared.
Predictive validity is concerned with how well a scale can forecast a future
criterion. To assess predictive validity, the researcher collects data on the scale at
one point in time and data on the criterion variables at a future time. For example,
attitudes towards how loyal customers feel to a particular brand could be used to
predict future repeat purchases of that brand. The predicted and actual purchases
are compared to assess the predictive validity of the attitudinal scale. Construct
validity addresses the question of what construct or characteristic the scale is, in
fact, measuring. When assessing construct validity, the researcher attempts to
answer theoretical questions about why the scale works and what deductions can
be made concerning the underlying theory. Thus, construct validity requires a
sound theory of the nature of the construct being measured and how it relates to
other constructs.
Construct validity is the most sophisti cated and difficult type of validity to
establish. Construct validity includes convergent, discriminant and nomological
validity. Convergent validity is the extent to which the scale correlates positively
with other measurements of the same construct. It i s not necessary that all these
measurements be obtained by using conventional scaling techniques. Discriminant
validity is the extent to which a measure does not correlate with other constructs
from which it is supposed to differ. It involves demonstrating a lack of correlation
among differing constructs. Nomological validity is the extent to which the scale
correlates in theoretically predicted ways with measures of different but related
constructs. A theoretical model is formulated that leads to further deductions, tests
and inferences. An example of construct validity can be evaluated in the following
example. A researcher seeks to provide evidence of construct validity in a multi -
item scale, designed to measure the concept of ‘self-image’. These findings would be
sought:■ High correlations with other scales designed to measure self-concepts and
with reported classifications by friends (convergent validity) ■ Low correlations with
unrelated constructs of brand loyalty and variety-seeking (discriminant validity) ■
74
Brands that are congruent with the individual’s self-concept are more preferred, as
postulated by the theory (nomological validity) ■ A high level of reliability.
Note that a high level of reliability was included as evidence of construct
validity in this example.
6.3.3 Scaling
In research we quite often face measurement problem (since we want a valid
measurement but may not obtain it), specially when the concepts to be measured
are complex and abstract and we do not possess the standardized measurement
tools. Alternatively, we can say that while measuring attitudes and opinions, we
face the problem of their valid measurement. Similar problem may be faced by a
researcher, of course in a lesser degree, while measuring physical or institutional
concepts. As such we should study some procedures which may enable us to
measure abstract concepts more accurately. This brings us to the study of scaling
techniques.
6.3.4 Meaning of Scaling
Scaling describes the procedures of assigning numbers to various degre es of
opinion, attitude and other concepts. This can be done in two ways viz., (i) making
a judgment about some characteristic of an individual and then placing him
directly on a scale that has been defined in terms of that characteristic and (ii)
constructing questionnaires in such a way that the score of individual's responses
assigns him a place on a scale. It may be stated here that a scale is a continuum,
consisting of the highest point (in terms of some characteristic e.g., preference,
favorableness, etc.) and the lowest point along with several intermediate points
between these two extreme points. These scale -point positions are so related to
each other that when the first point happens to be the highest point, the second
point indicates a higher degree in terms of a given characteristic as compared to the
third point and the third point indicates a higher degree as compared to the fourth
and so on. Numbers for measuring the distinctions of degree in the
attitudes/opinions are, thus, assigned to indivi duals corresponding to their scale-
positions. All this is better understood when we talk about scaling technique(s).
Hence the term 'scaling' is applied to the procedures for attempting to determine
quantitative measures of subjective abstract concepts. Sc aling has been defined as
a "procedure for the assignment of numbers (or other symbols) to a property of
objects in order to impart some of the characteristics of numbers to the properties
in question.
6.3.5 M easurement of attitudes in quantitative and qualitative research
There are a number of differences in the way attitudes are measured in
quantitative and qualitative research. In quantitative research you are able to
explore measure, determine the intensity and combine attitudes to different aspects
of an issue to arrive at one indicator that is reflective of the overall attitude. In
qualitative research, you can only explore the spread of attitudes and establish the
types of attitudes prevalent. In quantitative research you can ascertain the types of
attitudes people have in a community, how many people have a particular attitude
75
and what the intensity is of those attitudes. A number of techniques have been
developed to measure attitudes and their intensity in quantitative research, but
such techniques are lacking in qualitative research. This is mainly because in
qualitative research you do not make an attempt to measure or quantify. The
concept of attitudinal scales, therefore, is only prevalent in quantitative research.
6.3.6 Attitudinal scales in quantitative research
In quantitative research there are three scales which have been developed to
'measure' attitudes. Each of these scales is based upon different assumptions and
follows different procedures in their construction. As a beginner in research
methods it is important for you to understand these procedures and the
assumptions behind them so that you can make appropriate and accurate
interpretation of the findings. As you will see, it is not very easy to construct an
attitudinal scale. Out of the three scales, the Likert scale is the easiest to construct
and therefore is used far more.
6.4.7 Functions of attitudinal scales
If you want to find out the attitude of respondents towards an issue, you can
ask either a closed or an open-ended question. For example, let us say that you
want to ascertain the attitude of students in a class towards their lecturer and that
you have asked them to respond to the following question: 'What is your attitude
towards your lecturer?' If your question is open ended, it invites each respondent to
describe the attitude that s/he holds towards the lecturer. If you have framed a
closed question, with categories such as 'extremely positive', 'positive', 'uncertain',
'negative' and 'extremely negative', this guides the respondents to select a category
that best describes their attitude. This type of questioning, whether framed
descriptively or in a categorical form, elicits an overall attitude towards the lecturer.
While ascertaining the overall attitude may be sufficient in some situations, in
many others, where the purpose of attitudinal questioning is to develop strategies
for improving a service or intervention, or to formulate policy, eliciting attitudes on
various aspects of the issue under study is required.
But as you know, every issue, including that of the attitude of students
towards their lecturers, has many aspects. For example, the attitude of the
members of a community towards the provision of a particular service comprises
their attitude towards the need for the service, its manner of delivery, its location,
the physical facilities provided to users, the behaviour of the staff, the competence
of the staff, the effectiveness and efficiency of the service, and so on. Similarly,
other examples - such as the attitude of employees towards the management of
their organisation, the attitude of employees towards occupational redeployment
and redundancy, the attitude of nurses towards death and dying, the attitude of
consumers towards a particular product, the attitude of students towards a
lecturer, or the attitude of staff towards the strategic plan for their organisation -
can be broken down in the same manner.
76
the attitudes towards various aspects of an issue, though the degree of difficulty in
following the procedure for these scales varies from scale to scale.
6.4.6 Types of attitudinal scale
Likert’s Scale
Likert, is extremely popular for measuring attitudes, because, the method is
simple to administer. With the Likert scale, the respondents indicate their own
attitudes by checking how strongly they agree or disagree with carefully w orded
statements that range from very positive to very negative towards the attitudinal
object. Respondents generally choose from five alternatives (say strongly agree,
agree, neither agree nor disagree, disagree, strongly disagree). A Likert scale may
include a number of items or statements. Disadvantage of Likert Scale is that it
takes longer time to complete than other itemised rating scales because
respondents have to read each statement. Despite the above disadvantages, this
scale has several advantages. It is easy to construct, administer and use.
Semantic Differential Scale
This is a seven point rating scale with end points associated with bipolar labels
(such as good and bad, complex and simple) that have semantic meaning. It can be
used to find whether a respondent has a positive or negative attitude towards an
object. It has been widely used in comparing brands, products and company
images. It has also been used to develop advertising and promotion strategies and
in a new product development study.
Thurstone scale
Thurstone's method of pair comparisons can be considered a prototype of a
normal distribution-based method for scaling-dominance matrices. Even though
the theory behind this method is quite complex the algorithm itself is
straightforward. A Thurstone scale has a number of statements to which the
respondent is asked to agree or disagree.
There are three types of scale that Thurstone described:
Equal-appearing intervals method
Successive intervals method
Paired comparisons method
Multi-Dimensional Scaling
Multidimensional scaling (MDS) is a means of visualizing the level of similarity
of individual cases of a dataset. It refers to a set of related ordination techniques
used in information visualization, in particular to display the information contained
in a distance matrix.
6.4 REVISION POINTS
1. Measurement is a relatively complex and demanding task, specially so
when it concerns qualitative or abstract phenomena
2. Measurement is a process of mapping aspects of a domain onto other
aspects of a range according to some rule of correspondence
78
80
LESSON – 7
An expectant pause: If the interviewer believes the respondent has more to say,
the “silent probe:’ accompanied by an expectant look, may motivate the respondent
to gather his or her thoughts and give a complete response Of course, the
interviewer must be sensitive to the respondent so that the silent probe does not
become an embarrassed silence.
Repetition of the Respondent’s Reply: Sometimes the interviewer may repeat
the verbatim of the respondent. This may help the respondent to expand the
answer.
2. Recording the Responses
The analyst who fails to instruct fieldworkers in the techniques of recording
answers for one study rarely forgets to do so in the second study. Although the
concept of recording an answer seems extremely simple, mistakes c an be made in
the recording phase of the research. All fieldworkers should use the same
mechanics of recording.
Example: It may appear insignificant to the interviewer whether she uses a pen
or pencil, but to the editor who must erase and rewrite illegibl e words, using a
pencil is extremely important.
The rules for recording responses to closed questionnaires vary with the
specific questionnaire. The general rule, however, is to place a check in the box that
correctly reflects the respondent’s answer. All too often interviewers don’t bother
recording the answer to a filter question because they believe that the subsequent
answer will make the answer to the filter question obvious. However, editors and
coders do not know how the respondent actually answered a question.
The general instruction for recording answers to open -ended-response
questions is to record the answer verbatim, a task that is difficult for most people.
Inexperienced interviewers should be given the opportunity to practice
The Interviewer’s Manual of the Survey Research Center provides the
instructions on the recording of interviews. Some of its suggestions
Recording answers to open-ended-response questions follow:
Record the responses during the interview.
Use the respondent’s own words.
Do not summarize or paraphrase the respondent’s answer.
Include everything that pertains to the question objectives.
Include all of your probes;
The basics of Effective Interviewing
Interviewing is a skilled occupation; not everyone can do it, and even few can
do it extremely well. A good interviewer observes the following principles:
1. Have integrity and be honest.
2. Have patience and tact.
84
but not the name of the sponsoring bank. To train the interviewers about the
questionnaire, a field supervisor conducts an interview with another field
supervisor who acts as a respondent. The trainees observe the interviewing process,
after which they each interview and record the responses of another field
supervisor. Additional instructions are given to the trainees after the practice
interview.
Training to Avoid Procedural Errors in Sample Selection
The briefing session also covers the sampling procedure. A number of research
projects allow the interviewer to be at least partially responsible for selection of the
sample. When the fieldworker has some discretion in the selection of respondents,
the potential for selection bias exists. This is obvious in the case of quota sampling,
but less obvious in other cases.
Example: in probability sampling where every nth house is selected, the
fieldworker uses his or her discretion in identifying housing units. Avoiding
selection error may not be as simple as it sounds.
Example: in an older, exclusive neighborhood, a mansion’s coach house or
servants’ quarters may have been converted into an apartment that should be
identified as a housing unit. This type of dwelling and other unusual housing units
(apartments with alley entrances only, lake cottages, rooming houses) may be
overlooked, giving rise to selection error. Errors may also occur in the selection of
random-digit dialing samples Considerable effort should be expended in training
and supervisory control to minimize these errors.
The activities involved in collecting data in the field may be performed by the
organization needing information, by research suppliers, or by third -party field
service organizations. Proper execution of fieldwork is essential for producing
research results without substantial error.
Proper control of fieldwork begins with interviewer selection. Fieldworkers
should generally be healthy, outgoing, and well groomed. New fieldworkers must be
trained in opening the interview, aski ng the questions, probing for additional
information, recording the responses, and terminating the interview. Experienced
fieldworkers are briefed for each new project so that they are familiar with its
specific requirements. A particular concern of the briefing session is reminding
fieldworkers to adhere closely to the prescribed sampling procedures.
Careful supervision of fieldworkers is also necessary. Supervisors gather and
edit questionnaires each day. They check to see that field procedures are properly
followed and that interviews are on schedule. They also check to be sure that the
proper sampling units are used and that the proper people are responding in the
study. Finally, supervisors check for interviewer cheating and verify a portion of the
interviews by reinterviewing a certain percentage of each fieldworker’s respondents.
7.4 REVISION POINTS
1. It is important that the research administrator select capable people who
may be entrusted to collect the data
86
LESSON – 8
SAMPLING
8.1 INTRODUCTION
Sampling is the statistical process of selecting a subset (called a "sample") of a
population of interest for purposes of making observations and statistical
inferences about that population. Social science research is generally about
inferring patterns of behaviors within specific populations. We cannot study entire
populations because of feasibility and cost constraints, and hence, we must select a
representative sample from the population of interest for observati on and analysis.
It is extremely important to choose a sample that is truly representative of the
population so that the inferences derived from the sample can be generalized back
to the population of interest.
8.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Sampling
Sampling error
Sampling technique
Sampling size estimation
8.3 CONTENTS
8.3.1 Sampling Issues
8.3.2 The sampling process
8.3.3 Determination of sample size
8.3.4 mImportance of sampling design and sample size
8.3.1 Sampling Issues
One of the questions researchers face when developing a research plan is who
will be chosen to be research participants. The word ‘population’ is commonly used
to define everyone of interest who could be possibly included in a research stud y.
Researchers may define a population by geographic area. In addition, they may also
define a population using such demographic data as age, gender, income, or
ethnicity. Because marketing promotion is often targeted based on psychographic
segmentation, researchers may also define a population based on interests, values
or lifestyles. Product usage can also be a means for defining a population, such as
nonusers, occasional users and frequent users. Of course these variables can also
be used in combination. The resulting
This fact might be how many consumers would buy a product, what type of
new promotion would work best, the effect of pricing on purchase behavior, or the
best store in which to distribute a product. Of course if a 100 per cent accurate
answer is needed, researchers must ask every person who is included in a
population.
89
The sampling process comprises of several stage. The first stage is defining the
target population. A population can be defined as all people or items (unit of
analysis) with the characteristics that one wishes to study. The unit of analysis
may be a person, group, organization, country, object, or any other entity that you
wish to draw scientific inferences about. Sometimes the population is obvious. For
example, if a manufacturer wants to determine whether finished goods
manufactured at a production line meets certain quality requirements or must be
scrapped and reworked, then the population consists of the entire set of finished
goods manufactured at that production facility. At other times, the target
population may be a little harder to understand. If you wish to identify the primary
91
drivers of academic learning among high school students, then what is your target
population: high school students, their teachers, school principals, or parents? The
right answer in this case is high school students, because you are interested in
their performance, not the performance of their teachers, parents, or schools.
Likewise, if you wish to analyze the behavior of roulette wheels to identify biased
wheels, your population of interest is not different observatio ns from a single
roulette wheel, but different roulette wheels (i.e., their behavior over an infinite set
of wheels).
The second step in the sampling process is to choose a sampling frame. This
is an accessible section of the target population (usually a l ist with contact
information) from where a sample can be drawn. If your target population is
professional employees at work, because you cannot access all professional
employees around the world, a more realistic sampling frame will be employee lists
of one or two local companies that are willing to participate in your study. If your
target population is organizations, then the Fortune 500 list of firms or the
Standard & Poor's (S&P) list of firms registered with the New York Stock exchange
may be acceptable sampling frames.
Note that sampling frames may not entirely be representative of the population
at large, and if so, inferences derived by such a sample may not be generalizable to
the population. For instance, if your target population is organizational employees
at large (e.g., you wish to study employee self-esteem in this population) and your
sampling frame is employees at automotive companies in the American Midwest,
findings from such groups may not even be generalizable to the American workforce
at large, let alone the global workplace. This is because the American auto industry
has been under severe competitive pressures for the last 50 years and has seen
numerous episodes of reorganization and downsizing, possibly resulting in low
employee morale and self-esteem. Furthermore, the majority of the American
workforce is employed in service industries or in small businesses, and not in
automotive industry. Hence, a sample of American auto industry employees is not
particularly representative of the American workforce. Likewise, the Fortune 500
list includes the 500 largest American enterprises, which is not representative of all
American firms in general, most of which are medium and small -sized firms rather
than large firms, and is therefore, a biased sampling frame. In contrast, the S&P
list will allow you to select large, medium, and/or small companies, depending on
whether you use the S&P large-cap, mid-cap, or small-cap lists, but includes
publicly traded firms (and not private firms) and hence stil l biased. Also note that
the population from which a sample is drawn may not necessarily be the same as
the population about which we actually want information. For example, if a
researcher wants to the success rate of a new "quit smoking" program, then the
target population is the universe of smokers who had access to this program, which
may be an unknown population. Hence, the researcher may sample patients
arriving at a local medical facility for smoking cessation treatment, some of whom
92
may not have had exposure to this particular "quit smoking" program, in which
case, the sampling frame does not correspond to the population of interest.
The last step in sampling is choosing a sample from the sampling frame
using a well-defined sampling technique. Sampling techniques can be grouped into
two broad categories: probability (random) sampling and non -probability sampling.
Probability sampling is ideal if generalizability of results is important for your
study, but there may be unique circumstances where non -probability sampling can
also be justified. These techniques are discussed below.
PROBABILITY SAMPLING
Probability sampling is a technique in which every unit in the population has a
chance (non-zero probability) of being selected in the sample, and this chance can
be accurately determined. Sample statistics thus produced, such as sample mean
or standard deviation, are unbiased estimates of population parameters, as long as
the sampled units are weighted according to their probability of selection. All
probability sampling have two attributes in common: (1) every unit in the
population has a known non-zero probability of being sampled, and (2) the
sampling procedure involves random selection at some point. The different types of
probability sampling techniques include:
Simple random sampling: In this technique, all possible subsets of a population
(more accurately, of a sampling frame) are given an equal probability of being
selected. The probability of selecting any set of n units out of a total of N units in a
sampling frame is WC„. Hence, sample statistics are unbiased estimates of
population parameters, without any weighting. Simple random sampling involves
randomly selecting respondents from a sampling frame, but with large sampling
frames, usually a table of random numbers or a computerized random number
generator is used. For instance, if you wish to select 200 firms to survey from a list
of 1000 firms, if this list is entered into a spreadsheet like Excel, you can use
Excel's RAND() function to generate random numbers for each of the 1000 clients
on that list. Next, you sort the list in increasing order of their corresponding
random number, and select the first 200 clients on that sorted list. This is the
simplest of all probability sampling techniques; however, the simplicity is also the
strength of this technique. Because the sampling frame is not subdivided or
partitioned, the sample is unbiased and the inferences are most generalizable
amongst all probability sampling techniques.
Systematic sampling: In this technique, the sampling frame is ordered
according to some criteria and elements are selected at regular intervals through
that ordered list. Systematic sampling involves a random start and then proceeds
with the selection of every k th element from that point onwards, where k = N/n,
where k is the ratio of sampling frame size N and the desired sample size n, and is
formally called the sampling ratio. It is important that the starting point is not
automatically the first in the list, but is instead ran domly chosen from within the
first k elements on the list. In our previous example of selecting 200 firms from a
93
list of 1000 firms, you can sort the 1000 firms in increasing (or decreasing) order of
their size (i.e., employee count or annual revenues), randomly select one of the first
five firms on the sorted list, and then select every fifth firm on the list. This process
will ensure that there is no overrepresentation of large or small firms in your
sample, but rather that firms of all sizes are generall y uniformly represented, as it
is in your sampling frame. In other words, the sample is representative of the
population, at least on the basis of the sorting criterion.
Stratified sampling: In stratified sampling, the sampling frame is divided into
homogeneous and non-overlapping subgroups (called "strata"), and a simple
random sample is drawn within each subgroup. In the previous example of
selecting 200 firms from a list of 1000 firms, you can start by categorizing the firms
based on their size as large (more than 500 employees), medium (between 50 and
500 employees), and small (less than 50 employees). You can then randomly select
67 firms from each subgroup to make up your sample of 200 firms. However, since
there are many more small firms in a sampling frame than large firms, having an
equal number of small, medium, and large firms will make the sample less
representative of the population (i.e., biased in favor of large firms that are fewer in
number in the target population). This is called non-proportional stratified sampling
because the proportion of sample within each subgroup does not reflect the
proportions in the sampling frame (or the population of interest), and the smaller
subgroup (large-sized firms) is over-sampled. An alternative technique will be to
select subgroup samples in proportion to their size in the population. For instance,
if there are 100 large firms, 300 mid-sized firms, and 600 small firms, you can
sample 20 firms from the "large" group, 60 from the "medium" group and 120 from
the "small" group. In this case, the proportional distribution of firms in the
population is retained in the sample, and hence this technique is called proportional
stratified sampling. Note that the non-proportional approach is particularly effective
in representing small subgroups, such as large -sized firms, and is not necessarily
less representative of the population compared to the proportional approach, as
long as the findings of the non-proportional approach is weighted in accordance to
a subgroup's proportion in the overall population.
Cluster sampling: If you have a population dispersed over a wide geographic
region, it may not be feasible to conduct a simple random sampling of the entire
population. In such case, it may be reasonable to divide the population into
"clusters" (usually along geographic boundaries), randomly sample a few clusters,
and measure all units within that cluster. For instance, if you wish to sample city
governments in the state of New York, rather than travel all over the state to
interview key city officials (as you may have to do with a simple random sample),
you can cluster these governments based on their counties, randomly select a set of
three counties, and then interview officials from every official in those counties.
However, depending on between-cluster differences, the variability of sample
estimates in a cluster sample will generally be higher than that of a simple random
94
sample, and hence the results are less generalizable to the population than those
obtained from simple random samples.
Matched-pairs sampling: Sometimes, researchers may want to compare two
subgroups within one population based on a specific criterion. For instance, why
are some firms consistently more profitable than other firms? To conduct such a
study, you would have to categorize a sampling frame of firms into "high profitable"
firms and "low profitable firms" based on gross margins, earnings per share, or
some other measure of profitability. You would then select a simple random sample
of firms in one subgroup, and match each firm in this group with a firm in the
second subgroup, based on its size, industry segment, and/or other matching
criteria. Now, you have two matched samples of high-profitability and low-
profitability firms that you can study in greater detail. Such matched-pairs
sampling technique is often an ideal way of understanding bipolar differences
between different subgroups within a given population.
Multi-stage sampling: The probability sampling techniques described
previously are all examples of single-stage sampling techniques. Depending on your
sampling needs, you may combine these single -stage techniques to conduct multi-
stage sampling. For instance, you can stratify a list of businesses based on firm
size, and then conduct systematic sampling within each stratum. This is a two-
stage combination of stratified and systematic sampling. Likewise, you can start
with a cluster of school districts in the state of New York, and within each cluster,
select a simple random sample of school s; within each school, select a simple
random sample of grade levels; and within each grade level, select a simple random
sample of students for study. In this case, you have a four-stage sampling process
consisting of cluster and simple random sampling.
NON-PROBABILITY SAMPLING
Nonprobability sampling is a sampling technique in which some units of the
population have zero chance of selection or where the probability of selection
cannot be accurately determined. Typically, units are selected based on certai n
non-random criteria, such as quota or convenience. Because selection is non -
random, nonprobability sampling does not allow the estimation of sampling errors,
and may be subjected to a sampling bias. Therefore, information from a sample
cannot be generalized back to the population. Types of non -probability sampling
techniques include:
Convenience sampling: Also called accidental or opportunity sampling, this is
a technique in which a sample is drawn from that part of the population that is
close to hand, readily available, or convenient. For instance, if you stand outside a
shopping center and hand out questionnaire surveys to people or interview them as
they walk in, the sample of respondents you will obtain will be a convenience
sample. This is a non-probability sample because you are systematically excluding
all people who shop at other shopping centers. The opinions that you would get
from your chosen sample may reflect the unique characteristics of this shopping
95
center such as the nature of its stores (e.g., high end-stores will attract a more
affluent demographic), the demographic profile of its patrons, or its location (e.g., a
shopping center close to a university will attract primarily university students with
unique purchase habits), and therefore may not be representative of the opinions of
the shopper population at large. Hence, the scientific generalizability of such
observations will be very limited. Other examples of convenience sampling are
sampling students registered in a certain class or sampling patients arriving at a
certain medical clinic. This type of sampling is most useful for pilot testing, where
the goal is instrument testing or measurement validation rather than obtaining
generalizable inferences.
Quota sampling: In this technique, the population is segmented into mutually-
exclusive subgroups (just as in stratified sampling), and then a non -random set of
observations is chosen from each subgroup to meet a predefined quota. In
proportional quota sampling, the proportion of respondents in each subgroup
should match that of the population. For instance, if the American population
consists of 70% Caucasians, 15% Hispanic -Americans, and 13% African-
Americans, and you wish to understand their voting preferences in an sample of 98
people, you can stand outside a shopping center and ask people their voting
preferences. But you will have to stop asking Hispanic -looking people when you
have 15 responses from that subgroup (or African -Americans when you have 13
responses) even as you continue sampling other ethnic groups, so that the ethnic
composition of your sample matches that of the general American population. Non-
proportional quota sampling is less restrictive in that you don't have to achieve a
proportional representation, but perhaps meet a minimum size in each subgroup.
In this case, you may decide to have 50 respondents from each of the three ethnic
subgroups (Caucasians, Hispanic-Americans, and African-Americans), and stop
when your quota for each subgroup is reached. Neither type of quota sampling will
be representative of the American population, since depending on whether your
study was conducted in a shopping center in New York or Kansas, your results may
be entirely different. The non-proportional technique is even less representative of
the population but may be useful in that it allows capturing the opinions of small
and underrepresented groups through oversampling.
Expert sampling: This is a technique where respondents are chosen in a non -
random manner based on their expertise on the phenomenon being studied. For
instance, in order to understand the impacts of a new governmental policy such as
the Sarbanes-Oxley Act, you can sample an group of corporate accountants who
are familiar with this act. The advantage of this approach is that since experts tend
to be more familiar with the subject matter than non experts, opinions from a
sample of experts are more credible than a sample that includes both experts and
non-experts, although the findings are still not generalizable to the overall
population at large.
Snowball sampling: In snowball sampling, you start by identifying a few
respondents that match the criteria for inclusion in your study, and then ask them
96
to recommend others they know who also meet your selection criteria. For instance,
if you wish to survey computer network administrators and you know of only one or
two such people, you can start with them and ask them to recommend others who
also do network administration. Although this method hardly leads to
representative samples, it may sometimes be the only way to reach hard-to-reach
populations or when no sampling frame is available.
8.4 Determination of Sample Size
Determination of appropriate sample size is crucial part of any Business
Research. The decision on proper sample size tremendously requires the use of
statistical theory. When a business research report is been evaluated, the
evaluation start with the question of How big the Sample size?
Having discussed various sampling designs it is important to focus the
attention on Sample Size. Suppose we select a sample size of 30 from the
population of 3000 through a simple random sampling procedure. Will we able to
generalize the findings to the population with confidence? So in this case what is
the sample size that would be required to carry out the research?
It is the known fact that larger the sample size, the more accurate the research
is. In fact this is the fact based on the statistics. According to this fact, increasing
the sample size decreases the width of the confi dence interval at a given confidence
level. When the Standard deviation of the population is unknown, a confidence
interval is calculated by using the formula
Confidence Interval µ = X ± KSX
S
Where as x = S / √n
In sum, choosing the appropriate sampling plan is one of the important
research design decisions the researcher has to make. The choice of a specific
design will depend broadly on the goal of research, the characteristics of the
population, and considerations of cost.
Issues of Precision and Confidence in determining Sample size
We now need to focus attention on the second aspect of the sampling design
issue—the sample size. Suppose we select 30 people from a population of 3,000
through a simple random sampling procedure. Will we be able to generalize our
findings to the population with confidence? What is the sample size that would be
required to make reasonably precise generalizations with confidence? What do
precision and confidence mean?
A reliable and valid sample should enable us to generalize the findings from
the sample to the population under investigation. No sample statistic (X, for
instance) is going to be exactly the same as the population parameter (S X), no
matter how sophisticated the probability sampling design is. Remember that the
very reason for a probability design is to increase the probability that the sample
statistics will be as close as possible to the population parameters!
97
Precision
Precision refers to how close our estimate is to the true population
characteristic. Usually, we would estimate the population parameter to fall within a
range, based on the sample estimate.
Example: From a study of a simple random sample of 50 of the total 300
employees in a workshop, we find that the average daily production rate per person
is 50 pieces of a particular product (X = 50). We might then (by doing certain
calculations, as we shall see later) be able to say that the true average daily
production of the product (X) would lie anywhe re between 40 and 60 for the
population of employees in the workshop. In saying this, we offer an interval
estimate, within which we expect the true population mean production to be (µ = 50
± 10). The narrower this interval, the greater the precision. For instance, if we are
able to estimate that the population mean would fall anywhere between 45 and 55
pieces of production (µ = 50 ± 5) rather than 40 and 60 (µ = 50 ± 10), then we would
have more precision. That is, we would now estimate the mean to lie within a
narrower range, which in turn means that we estimate with greater exactitude or
precision.
Precision is a function of the range of variability in the sampling distribution of
the sample mean. That is, if we take a number of different samples from a
population, and take the mean of each of these, we will usually find that they are
all different, are normally distributed, and have a dispersion associated with them.
Even if we take only one sample of 30 subjects from the population, we will still be
able to estimate the variability of the sampling distribution of the sample mean.
This variability is called the standard error, denoted by .S’. The standard error is
calculated by the following formula:
S
x = S / √n
Where S is the standard deviation of the sample, n is the sample size, and S X
indicates the standard error or the extent of precision offered by the sample.
In sum, the closer we want our sample results to reflect the population
characteristics, the greater will be the precision we would aim at. The greater the
precision required, the larger is the sample size needed, especially when the
variability in the population itself is large.
Confidence
Whereas precision denotes how close we estimate the population parameter
based on the sample statistic, confidence denotes how certain we are that our
estimates will really hold true for the population. In the previous example of
production rate, we know we are more precise when we estimate the true mean
production (µ) to fall somewhere between 45 and 55 pieces, than somewhere
between 40 and 60.
98
In essence, confidence reflects the level of certainty with which we can state
that our estimates of the population parameters, based on our sample statistics will
hold true. The level of confidence can range from 0 to 100% . A 95% confidence is
the conventionally accepted level for most business research, most commonly
expressed by denoting the significance level as p = .05. In other words, we say that
at least 95 times out of 100, our estimate will reflect the true population
characteristic,
In sum, the sample size, n, is a function of
1. The variability in the population
2. Precision or accuracy needed
3. Confidence level desired
4. Type of sampling plan used—for example, sample random sampling versus
stratified random sampling
1. thus becomes necessary for researchers to consider at least four points
while making decisions on the sample size needed to do the research
2. Much precision is really needed in estimating the population characteristics
interest—that is, what is the margin of allowable error?
3. How much confidence is really needed—that is, how much chance can we
take of making errors in estimating the population parameters?
4. To what extent is there variability in the population on the characteristic s
investigated?
5. What is the cost—benefit analysis of increasing the sample size?
Determining the Sample Size
Now that we are aware of the fact that the sample size is governed by the
extent of precision and confidence desired, how do we determine the sample size
required for our research? The procedure can be illustrated through an example.
Suppose a manager wants to be 95% confident that the expected monthly
withdrawals in a bank will be within a confidence interval of ± $500. Example of a
sample of clients indicates that the average withdrawals made by them have a
standard deviation of $3,500. What would be the sample size needed in this case?
We noted earlier that the population mean can be estimated by using the
formula:
S
µ= X ± K x
Since the confidence level needed here is 95% , the applicable K value is 1.96 (t
table). The interval estimate of ± $500 will have to encompass a dispersion of (1.96
x standard error). That is,
S
500 = 1.96 × x
99
S
500/1.96 = 255.10
x
We already know that
S
x = S / √n
255.10= 3500/ √n
n = 188
The sample size needed in the above was 188. Let us say that this bank has
the total clientele of only 185. This means we cannot sample 188 clients. We can in
this case apply the correction formula and see what sample size would be needed to
have the same level of precision and confidence given the fact that we have a total
of only 185 clients. The correction formula is as follows:
S
x = S/√n × √N – n /N-1
Where N is the total number of elements in the population, n is the sample
size to be estimated, Sx is the standard error of estimate of the mean, and S is the
standard deviation of the sample mean.
Applying the correlation formula, we fin d that
255.10 = 3500/√n × √185-n/184
n = 94
We would now sample 94 of the total 185 clients.
To understand the impact of precision and/or confidence on the sample size,
let us try changing the confidence level required in the bank withdrawal exercise
which needed a sample size of 188 for a confidence level of 95% . Let us say that the
bank manager now wants to be 99% sure that the expected monthly withdrawals
will be within the interval of ±$500. What will be the sample size now needed?
S
will now be 500/2.576 = 194.099
x
194.099 = 3500/√n
n = 325
The sample has now to be increased 1.73 times (from 188 to 325) to increase
the confidence level from 95% to 99% . It is hence a good idea to think through how
much precision and confidence one really needs, before determining the sample size
for the research project.
So far we have discussed sample size in the context of precision and
confidence with respect to one variable only. However, in research, the theoretical
framework has several variables of interest, and the question ari ses how one should
come up with a sample size when all the factors are taken into account.
100
Krejcie and Morgan (1970) greatly simplified size decision by providing a table
that ensures a good decision model. The Table provides that generalized scientific
guideline for sample size decisions. The interested Student is advised to read
Krejcie and Morgan (1970) as well as Cohen (1969) for decisions on sample size.
Table on Sample Size for a given Population Size
N S N S N S
10 10 220 140 1200 291
15 14 230 144 1300 297
20 19 240 148 1400 302
25 24 250 152 1500 306
30 28 260 155 1600 310
35 32 270 159 1700 313
40 36 280 162 1800 317
45 40 290 165 1900 320
50 44 300 169 2000 322
55 48 320 175 2200 327
60 52 340 181 2400 331
65 56 360 186 2600 335
70 59 380 191 2800 338
75 63 400 196 3000 341
80 66 420 201 3500 346
85 70 440 205 4000 351
90 73 460 210 4500 354
95 76 480 214 5000 357
100 80 500 217 6000 361
110 86 550 226 7000 364
120 92 600 234 8000 367
130 97 650 242 9000 368
140 103 700 248 10000 370
150 108 750 254 15000 375
160 113 800 260 20000 377
170 118 850 265 30000 379
180 123 900 269 40000 380
190 127 950 274 50000 381
200 132 1000 278 75000 382
210 136 1100 285 100000 384
8.5 Importance of Sampling Design and Sample Size
It is now possible to see how both sampling design and the sample size are
important to establish the representativeness of the sample for generalizability. If
the appropriate sampling design is not used, a large sample size will not, in itself,
allow the findings to be generalized to the population. Likewise, unless the sample
101
size is adequate for the desired level of precision and confidence, no sampling
design, however sophisticated, can be useful to the researcher in meeting the
objectives of the study.
Hence, sampling decisions should consider both the sampling design and the
sample size. Too large a sample size, however (say, over 500) could also become a
problem in as much as we would be prone to committing Type II errors. Hence,
neither too large nor too small sample sizes help research projects.
Roscoe (1975) proposes the following rules of thumb for determining sample
size
1. Sample sizes larger than 30 and less than 500 are appropriate for most
research.
2. Where samples are to be broken into sub samples; (male/females, juniors/
seniors, etc.), a minimum sample size of 30 for each category is necessary.
3. In multivariate research (including multiple regression analyses), the
sample size should be several times (preferably 10 times or more) as large
as the number of variables in the study.
4. For simple experimental research with tight experimental controls
(matched pairs, etc.), successful research is possible with samples as small
as 10 to 20 in size.
8.4 REVISION POINTS
1. Sampling is the statistical process of selecting a subset (called a "sample") of
a population of interest for purposes of making observations and statistical
inferences about that population.
2. The word ‘population’ is commonly used to define everyone of interest who
could be possibly included in a research study.
3. Selection errors occur when the correct population has been chosen, but the
sample taken from the population is not representative of the entire
population.
4. A population can be defined as all people or items (unit of analysis) with the
characteristics that one wishes to study.
5. A sampling frame is an accessible section of the target population (usually a
list with contact information) from where a sample can be drawn.
6. Probability sampling is a technique in which every unit in the population
has a chance (non-zero probability) of being selected in the sample, and this
chance can be accurately determined.
7. Nonprobability sampling is a sampling technique in which some units of the
population have zero chance of selection or where the probability of selection
cannot be accurately determined
102
104
LESSON – 9
HYPOTHESIS TESTING
9.1 INTRODUCTION
Statistical hypothesis is an assumption about a population parameter. This
assumption may or may not be true.
The best way to determine whether a statistical hypothesis is true is to
examine the entire population. Since this is often impractical, researchers typically
examine a random sample from the population. If the sample data are consistent
with the statistical hypothesis, the hypothesis is accepted; if not, the hypothesis is
rejected.
There are two types of statistical hypotheses.
Null hypothesis. The null hypothesis is usually the hypothesis that sample
observations result purely from chance effects.
Alternative hypothesis. The alternative hypothesis is the hypothesis that
sample observations are influenced by some non -random cause.
For example, suppose we wanted to determine whether a coin was fair and
balanced. A null hypothesis might be that half the flips would result in Heads and
half, in Tails. The alternative hypothesis might be that the number of Heads and
Tails would be very different. Suppose we flipped the coin 50 times, resulting in 40
Heads and 10 Tails. Given this result, we would be inclined to reject the null
hypothesis and accept the alternative hypothesis.
9.2 OBJECTIVES
After studying this lesson, you will be able to understand:
Hypothesis testing
Types of hypothesis
Meaning of probability value
Type I error and Type II error
Chi square test
9.3 CONTENTS
9.3.1 Hypothesis test
9.3.2 The precise meaning of the probability value
9.3.3 Statistical and practical significance
9.3.4 Type I and II errors
9.3.5 One and two tailed test
9.3.6 Confidence interval and hypothesis testing
9.3.7 Defining the decision rule and region of acceptance
9.3.8 Other considerations
9.3.9 Chi square test
105
the sample. There is also a more formal argument. A non significant effect might be
described as follows:
Although subjects in the drug group scored higher (M = 23) than did subjects
in the control group, (M = 20), the difference between means was not significant,
t(18) = 1.4, p = 0.179. It would not have been correct to say that there was no
difference between the performance of the two groups. There was a difference. It is
just that the difference was not large enough to rule out chance as an explanation
of the difference. It would also have been incorrect to imply that there is no
difference in the population. Be sure not to accept the null hypothesis.
9.3.2 The Precise Meaning of the Probability Value
There is often confusion about the precise meaning of the probability
computed in a significance test. As stated in Step 4 of the steps in hypothesis
testing, the null hypothesis (H0) is assumed to be true. The difference between the
statistic computed in the sample and the parameter specified by H0 is computed
and the probability of obtaining a difference this large or large is calculated. This
probability value is the probability of obtaining data as extreme or more extreme
than the current data (assuming H0 is true). It is not the probability of the null
hypothesis itself. Thus, if the probability value is 0.005, this does not mean that the
probability that the null hypothesis is true is .005. It means that the probability of
obtaining data as different or more different from the null hypothesis as those
obtained in the experiment is 0.005.
The inferential step to conclude that the null hypothesis is false goes as
follows: The data (or data more extreme) are very unlikely given that the null
hypothesis is true. This means that: (1) a very unlikely event occurred or (2) the
null hypothesis is false. The inference usually made is that the null hypothesis is
false.
To illustrate that the probability is not the probability of the hypothesis,
consider a test of a person who claims to be able to predict whether a coin will come
up heads or tails. One should take a rather skeptical attitude toward this claim and
require strong evidence to believe in its validity. The null hypothesis is that the
person can predict correctly half the time (H 0: π = 0.5). In the test, a coin is flipped
20 times and the person is correct 11 times. If the person has no special ability (H 0
is true), then the probability of being correct 11 or more times out of 20 is 0.41.
Would someone who was originally skeptical now believe that there is only a 0.41
chance that the null hypothesis is true? They almost certainly would not since they
probably originally thought H0 had a very high probability of being true (perhaps as
high as 0.9999). There is no logical reason for them to decrease their belief in the
validity of the null hypothesis since the outcome was perfectly consistent with the
null hypothesis.
The proper interpretation of the test is as follows: A person made a rather
extraordinary claim and should be able to provide strong evidence in support of the
109
claim if the claim is to believed. The test provided data consistent with the null
hypothesis that the person has no special ability since a person with no special
ability would be able to predict as well or better more than 40% of the time.
Therefore, there is no compelling reason to believe the extraordinary claim.
However, the test does not prove the person cannot predict better than chance; it
simply fails to provide evidence that he or she can. The probability that the null
hypothesis is true is not determined by the statistical analysis conducted as part of
hypothesis testing. Rather, the probability computed is the probability of obtaining
data as different or more different from the null hypothesis (given that the null
hypothesis is true) as the data actually obtained.
According to one view of hypothesis testing, the significance level should be
specified before any statistical calculations are performed. Then, when the
probability (p) is computed from a sign ificance test, it is compared with the
significance level. The null hypothesis is rejected if p is at or below the significance
level; it is not rejected if p is above the significance level. The degree to which p
ends up being above or below the significance level does not matter. The null
hypothesis either is or is not rejected at the previously stated significance level.
Thus, if an experimenter originally stated that he or she was using the 0.05
significance level and p was subsequently calculated to be 0.042, then the person
would reject the null hypothesis at the 0.05 level. If p had been 0.0001 instead of
0.042 then the null hypothesis would still be rejected at the 0.05 level. The
experimenter would not have any basis to be more confident that the nu ll
hypothesis was false with a p of 0.0001 than with a p of 0.041. Similarly, if the p
had been 0.051 then the experimenter would fail to reject the null hypothesis.
He or she would have no more basis to doubt the validity of the null
hypothesis than if p had been 0.482. The conclusion would be that the null
hypothesis could not be rejected at the 0.05 level. In short, this approach is to
specify the significance level in advance and use p only to determine whether or not
the null hypothesis can be rejected at the stated significance level.
Many statisticians and researchers find this approach to hypothesis testing
not only too rigid, but basically illogical. Who in their right mind would not have
more confidence that the null hypothesis is false with a p of 0.0001 then with a p of
0.042? The less likely the obtained results (or more extreme results) under the null
hypothesis, the more confident one should be that the null hypothesis is false. The
null hypothesis should not be rejected once and for all. The possibility that it was
falsely rejected is always present, and, all else being equal, the lower the p value,
the lower this possibility.
9.3.3 Statistical and Practical Significance
It is important not to confuse the confidence with which the null hypothe sis
can be rejected with size of the effect. To make this point concrete, consider a
researcher assigned the task of determining whether the video display used by
travel agents for booking airline reservations should be in color or in black and
110
white. Market research had shown that travel agencies were primarily concerned
with the speed with which reservations can be made. Therefore, the question was
whether color displays allow travel agents to book reservations faster. Market
research had also shown that in order to justify the higher price of color displays,
they must be faster by an average of at least 10 seconds per transaction. Fifty
subjects were tested with color displays and 50 subjects were tested with black and
white displays.
Subjects were slightly faster at making reservations on a color display (M =
504.7 seconds) than on a black and white display (M = 508.2) seconds. Although
the difference is small, it was statistically significant at the .05 significance level.
Box plots of the data are shown below.
difference of 3.4 or larger with a sample size of 100 is less likely than a sample
difference of 3.5 or larger with a sample size of 50.
The 95% confidence interval on the difference between means is:
-5.8 ≤ μcolor - μblack & white ≤ -0.9 and the 99% interval is:
-6.6 ≤ μcolor - μblack & white ≤ -0.1
Therefore, despite the finding of a "more significant" difference between means,
the experimenter can be even more certain that the color displays are only slightly
better than the black and white displays. The second experiment shows
conclusively that the difference is less than 10 seconds.
This example was used to illustrate the following points: (1) an effect that is
statistically significant is not necessarily large enough to be of practical significance
and (2) the smaller of two effects can be "more significant" than the larger. Be
careful how you interpret findings reported in the media. If you read that a
particular diet lowered cholesterol significantly, this does not necessarily mean that
the diet lowered cholesterol enough to be of any health value. It means that the
effect on cholesterol in the population is greater than zero.
9.3.4 Type I and II errors (1 of 2)
There are two kinds of errors that can be made in significance testing: (1) a
true null hypothesis can be incorrectly rejected and (2) a false null hypothesis can
fail to be rejected. The former error is called a Type I error and the latter error is
called a Type II error. These two types of errors are defined in the table.
True State of the Null Hypothesis
Statistical Decision
H0 True H0 False
Reject H0 Type I error Correct
Do not Reject H0 Correct Type II error
The probability of a Type I error is designated by the Greek letter alpha () and
is called the Type I error rate; the probability of a Type II error (the Type II error
rate) is designated by the Greek letter beta (ß). A Type II error is only an error in the
sense that an opportunity to reject the null hypothesis correctly was lo st. It is not
an error in the sense that an incorrect conclusion was drawn since no conclusion is
drawn when the null hypothesis is not rejected.
A Type I error, on the other hand, is an error in every sense of the word. A
conclusion is drawn that the null hypothesis is false when, in fact, it is true.
Therefore, Type I errors are generally considered more serious than Type II errors.
The probability of a Type I error (α) is called the significance level and is set by the
experimenter. There is a tradeoff between Type I and Type II errors. The more an
experimenter protects himself or herself against Type I errors by choosing a low
level, the greater the chance of a Type II error. Requiring very strong evidence to
reject the null hypothesis makes it very unlikely that a true null hypothesis will be
112
rejected. However, it increases the chance that a false null hypothesis will not be
rejected, thus lowering power. The Type I error rate is almost always set at .05 or at
.01, the latter being more conservative since it requires stronger evidence to reject
the null hypothesis at the .01 level then at the .05 level.
9.3.5 One- and Two-Tailed Tests
In the section on "Steps in hypothesis testing" the fourth step involves
calculating the probability that a statistic would differ as much or more from
parameter specified in the null hypothesis as does the statistic obtained in the
experiment. This statement implies that a difference in either direction would be
counted. That is, if the null hypothesis were:
H0: μ- μ = 0 and the value of the statistic M 1- M2 were +5, then the probability
of M1- M2 differing from zero by five or more (in either direction) would be
computed. In other words, probability value would be the probabi lity that either M1-
M2 ≥ 5 or M1- M2 ≤ -5. Assume that the figure shown below is the sampling
distribution of M 1- M2.
The figure shows that the probability of a value of +5 or more is 0.036 and that
the probability of a value of -5 or less is .036. Therefore the probability of a value
either greater than or equal to +5 or less than or equal to -5 is 0.036 + 0.036 =
0.072.
A probability computed considering differences in both directions is called a
"two-tailed" probability. The name makes sense since both tails of the sampling
distribution are considered. There are situations in which an experimenter is
concerned only with differences in one direction. For example, an experimenter may
be concerned with whether or not μ 1 - μ 2 is greater than zero. However, if μ 1 - μ 2 is
not greater than zero, the experimenter may not care whether it equals zero or is
less than zero. For instance, if a new drug treatment is developed, the main issue is
whether or not it is better than a placebo. If the treatment is not better than a
placebo, then it will not be used. It does not really matter whether or not it is worse
than the placebo.
When only one direction is of concern to an experimenter, then a "one -tailed"
test can be performed. If an experimenter were only concerned with whether or not
μ 1 - μ 2 is greater than zero, then the one-tailed test would involve calculating the
probability of obtaining a statistic as greater than the one obtained in the
experiment.
113
The shaded area in the figure is greater than five. The figure shows that the
one-tailed probability is 0.036.
It is easier to reject the null hypothesis with a one -tailed than with a two-tailed
test as long as the effect is in the specified direction. Therefore, one-tailed tests
have lower Type II error rates and more power than do two-tailed tests. In this
example, the one-tailed probability (0.036) is below the conventional significance
level of 0.05 whereas the two-tailed probability (0.072) is not. Probability values for
one-tailed tests are one half the value for two-tailed tests as long as the effect is in
the specified direction.
One-tailed and two-tailed tests have the same Type I error rate. One -tailed
tests are sometimes used when the experimenter predicts the direction of the effect
in advance. This use of one-tailed tests is questionable because the experimenter
can only reject the null hypothesis if the effect is in the predicted direction. If the
effect is in the other direction, then the null hypothesis cannot be rejected no
matter how strong the effect is. A skeptic might question whether the experimenter
would really fail to reject the null hypothesis if the effect were strong enough in the
wrong direction. Frequently the most intere sting aspect of an effect is that it runs
counter to expectations. Therefore, an experimenter who committed himself or
herself to ignoring effects in one direction may be forced to choose between ignoring
a potentially important finding and using the techn iques of statistical inference
dishonestly. One-tailed tests are not used frequently. Unless otherwise indicated, a
test should be assumed to be two-tailed.
9.3.6 Confidence Intervals & Hypothesis Testing
There is an extremely close relationship between c onfidence intervals and
hypothesis testing. When a 95% confidence interval is constructed, all values in the
interval are considered plausible values for the parameter being estimated. Values
outside the interval are rejected as relatively implausible. If the value of the
parameter specified by the null hypothesis is contained in the 95% interval then the
null hypothesis cannot be rejected at the 0.05 level. If the value specified by the
null hypothesis is not in the interval then the null hypothesis can be rejected at the
0.05 level. If a 99% confidence interval is constructed, then values outside the
interval are rejected at the 0.01 level.
114
Imagine a researcher wishing to test the null hypothesis that the mean time to
respond to an auditory signal is the same as the mean time to respond to a visual
signal. The null hypothesis therefore is:
μvisual - μauditory = 0.
Ten subjects were tested in the visual condition and their scores (in
milliseconds) were: 355, 421, 299, 460, 600, 580, 474, 511, 550, and 586.
Ten subjects were tested in the auditory condition and their scores were: 275,
320, 278, 360, 430, 520, 464, 311, 529, and 326.
The 95% confidence interval on the difference between means is:
9 ≤ μvisual - μauditory ≤ 196.
Therefore only values in the interval between 9 and 196 are retained as
plausible values for the difference between population means. Since zero, the value
specified by the null hypothesis, is not in the interval, the null hypothesis of no
difference between auditory and visual presentation can be rejected at the 0.05
level. The probability value for this example is 0.034. Any time the parameter
specified by a null hypothesis is not contained in the 95% confidence interval
estimating that parameter, the null hypothesis can be rejected at the 0.05 level or
less. Similarly, if the 99% interval does not contain the parameter then the null
hypothesis can be rejected at the 0.01 level. The null hypothesis is not rejected if
the parameter value specified by the null hypothesis is in the interval since the null
hypothesis would still be plausible.
However, since the null hypothesis would be only one of an infinite number of
values in the confidence interval, accepting the null hypothesis is not justified.
There are many arguments against accepting the null hypothesis when it is not
rejected. The null hypothesis is usually a hypothesis of no difference. Thus null
hypotheses such as:
μ1 - μ2 = 0
π1 - π2 = 0
in which the hypothesized value is zero are most common. When the
hypothesized value is zero then there is a simple relationship between hypothesis
testing and confidence intervals:
If the interval contains zero then the null hypothesis cannot be rejected at the
stated level of confidence. If the interval does not contain zero then the null
hypothesis can be rejected.
This is just a special case of the general rule stating that the null hypothesis
can be rejected if the interval does not contain the hypothesized value of the
parameter and cannot be rejected if the interval contains the hypothesize d value.
Since zero is contained in the interval, the null hypothesis that μ 1 - μ 2 = 0 cannot
be rejected at the 0 .05 level since zero is one of the plausible values of μ 1 - μ 2. The
115
interval contains both positive and negative numbers and therefore μ 1 may be
either larger or smaller than μ 2. None of the three possible relationships between μ 1
and μ 2:
μ 1 - μ 2 = 0,
μ 1 - μ 2 > 0, and
μ1 - μ2 < 0
can be ruled out. The data are very inconclusive. Whenever a significance test
fails to reject the null hypothesis, the direction of the effect (if there is one) is
unknown.
Now, consider the 95% confidence interval:
6 ≤ μ1 - μ2 ≤ 15.
Since zero is not in the interval, the null hypothesis that μ 1 - μ 2 = 0 can be
rejected at the 0.05 level. Moreover, since all the values in the interval are positive,
the direction of the effect can be inferred: μ 1 > μ 2.
Whenever a significance test rejects the null hypothe sis that a parameter is
zero, the confidence interval on that parameter will not contain zero. Therefore
either all the values in the interval will be positive or all the values in the interval
will be negative. In either case, the direction of the effect is known.
9.3.7 Define the Decision Rule and the Region of Acceptance
The decision rule consists of two parts: (1) a test statistic and (2) a range of
values, called the region of acceptance. The decision rule determines whether a null
hypothesis is accepted or rejected. If the test statistic falls within the region of
acceptance, the null hypothesis is accepted; otherwise, it is rejected.
We define the region of acceptance in such a way that the chance of making a
Type I error is equal to the significance level. Here is how that is done.
Given the significance level α , find the upper limit (UL) of the range of
acceptance. There are three possibilities, depending on the form of the null
hypothesis.
If the null hypothesis is μ < M: The upper limit of the region of
acceptance will be equal to the value for which the cumulative
probability of the sampling distribution is equal to one minus the
significance level. That is, P( x < UL ) = 1 - α .
If the null hypothesis is μ = M: The upper limit of the region of
acceptance will be equal to the value for which the cumulative
probability of the sampling distribution is equal to one minus the
significance level divided by 2. That is, P( x < UL ) = 1 - α/2 .
If the null hypothesis is μ > M: The upper limit of the region of
acceptance is equal to plus infinity.
116
In a similar way, we find the lower limit (LL) of the range of acceptance.
Again, there are three possibilities, depending on the form of the null
hypothesis.
If the null hypothesis is μ < M: The lower limit of the region of
acceptance is equal to minus infinity.
If the null hypothesis is μ = M: The lower limit of the region of
acceptance will be equal to the value for which the cumulative
probability of the sampling distribution is equal to the significance
level divided by 2. That is, P( x < LL ) = α/2 .
If the null hypothesis is μ > M: The lower limit of the region of
acceptance will be equal to the value for which the cumulative
probability of the sampling distribution is equal to the significance
level. That is, P( x < LL ) = α .
The region of acceptance is defined by the range between LL and UL.
Accept or Reject the Null Hypothesis
Once the region of acceptance is defined, the null hypothesis can be tested
against sample data. Simply compute the test statistic. In this case, the test
statistic is the sample mean. If the sample mean falls within the region of
acceptance, the null hypothesis is accepted; if not, it is rejected.
9.3.8 Other Considerations
When one tests a hypothesis in the real world, other issues may come into
play. Here are some suggestions that may be helpful.
You will need to make an assumption about the sampling distribution of
the mean score. If the sample is relatively large (i.e., greater than or equal
to 30), you can assume, based on the central limit theorem, that the
sampling distribution will be roughly normal. On the other hand, if the
sample size is small (less than 30) and if the population random variable is
approximately normally distributed (i.e., has a bell -shaped curve), you can
transform the mean score into a t score. The t score will have a (duh) t
distribution.
Assume that the mean of the sampling distribution is equal to the test
value M specified in the null hypothesis.
In some situations, you may need to compute the standard deviation of
the sampling distribution sx. If the standard deviation of the population σ
is known, then
sx = σ * sqrt[( 1/n ) - ( 1/N )]
where n is the sample size and N is the population size. On the other hand, if
the standard deviation of the population σ is unknown, then
sx = s * sqrt[( 1/n ) - ( 1/N )]
where s is the sample standard deviation.
117
Example 1
An inventor has developed a new, energy-efficient lawn mower engine. He
claims that the engine will run continuously for 5 hours (300 minutes) on a single
gallon of regular gasoline. Suppose a random sample of 50 e ngines is tested. The
engines run for an average of 295 minutes, with a standard deviation of 20
minutes. Test the null hypothesis that the mean run time is 300 minutes against
the alternative hypothesis that the mean run time is not 300 minutes. Use a 0.0 5
level of significance.
Solution
There are four steps in conducting a hypothesis test, as described in the
previous sections. We work through those steps below:
Formulate hypotheses. The first step is to state the null hypothesis and an
alternative hypothesis.
Null hypothesis: μ = 300 minutes
Define the decision rule. The decision rule consists of two parts: (1) a test
statistic and (2) a range of values, called the region of acceptance. We
already know that the test statistic is a sample mean equal to 108. All that
remains is to describe the region of acceptance; that is, to define the lower
limit and the upper limit of the region. Here is how that is done.
Specify the sampling distribution. Since the sample size is small (less
than 30), we assume that the sampling distribution of the mean
follows a t distribution.
Define the mean of the sampling distribution . We assume that the
mean of the sampling distribution is equal to the mean value that
appears in the null hypothesis - 110.
Find the lower limit of the region of acceptance. Given a one-tailed
hypothesis, the lower limit (LL) will be equal to the value for which
the cumulative probability of the sampling distribution is equal to the
significance level. That is, P( x < LL ) = α = 0.01. To find this lower
limit, we use the T Distribution Table. From table: cumulative
probability = 0.01, population mean = 110, standard deviation = 2.16,
and degrees of freedom = 20 - 1 = 19. The calculation tells us that the
sample mean is 104.51, given those inputs. This is the lower limit of
our region of acceptance.
Find the upper limit of the region of acceptance. Since we have a one-
tailed hypothesis in which the null hypothesis states that the IQ is
greater than 110, any big number is consistent with the null
hypothesis. Therefore, the upper limit is plus infinity.
Thus, we have determined that the range of acceptance is defined by the
values between 104.51 and plus infinity.
Accept or reject the null hypothesis. The sample mean in this example was
an IQ score of 108. This value falls within the region of acceptance.
Therefore, we cannot reject the null hypothesis that the average IQ score
of students at Bon Air Elementary is equal to 110.
Power of a Hypothesis Test
When we conduct a hypothesis test, we accept or reject the null hypothesis
based on sample data. Because of the random nature of sample data, our decision
can have four possible outcomes.
We may accept the null hypothesis when it is true. Thus, the decision is
correct.
We may reject the null hypothesis when it is true. This kind of incorrect
decision is called a Type I error.
We may reject the null hypothesis when it is false. Thus, the decision is
correct.
120
We may accept the null hypothesis when it is false. This kind of incorrect
decision is called a Type II error.
The probability of committing a Type I error is called the significance level and
is denoted by α. The probability of committing a Type II error is called Beta and is
denoted by β. The probability of not committing a Type II error is called the power of
the test.
How to Compute the Power of a Test
When a researcher designs a study to test a hypothesis, he/she should
compute the power of the test (i.e., the likelihood of avoiding a Type II error). Here is
how to do that.
Define the region of acceptance. (The process for defining the region of
acceptance is described in the previous three lessons. See, for example,
Hypothesis Tests of Mean Score, Hypothesis Tests of Proportion (Large
Sample), and Hypothesis Tests of Proportion (Small Sample).)
Specify the critical value. The critical value is an alternative to the value
specified in the null hypothesis. The difference between the critical value
and the value from the null hypothesis is called the effect size. That is, the
effect size is equal to the critical value minus the value from the null
hypothesis.
Compute power. Assume that the true population parameter is equal to
the critical value, rather than the value specified in the null hypothesis.
Based on that assumption, compute the probability that the sample
estimate of the population parameter will fall outside the region of
acceptance. That probability is the power of the test.
Example 1: Power of the Hypothesis Test of a Mean Score
Two inventors have developed a new, energy-efficient lawn mower engine. One
inventor says that the engine will run continuously for 5 hours (300 mi nutes) on a
single gallon of regular gasoline. Suppose a random sample of 50 engines is tested.
The engines run for an average of 295 minutes, with a standard deviation of 20
minutes. The inventor tests the null hypothesis that the mean run time is 300
minutes against the alternative hypothesis that the mean run time is not 300
minutes, using a 0.05 level of significance.
The other inventor says that the new engine will run continuously for only 290
minutes on a gallon of gasoline. Find the power of the te st to reject the null
hypothesis, if the second inventor is correct.
Solution
The steps required to compute power are presented below.
Define the region of acceptance. In a previous lesson, we showed that the
region of acceptance for this problem consists of the values between
294.45 and 305.55
121
Specify the critical value. The null hypothesis tests the hypothesis that the
run time of the engine is 300 minutes. We are interested in determining
the probability that the hypothesis test will reject the null hypothesis, if
the true run time is actually 290 minutes. Therefore, the critical value is
290. (Another way to express the critical value is through effect size. The
effect size is equal to the critical value minus the hypothesized value.
Thus, effect size is equal to 290 - 300 or -10.)
Compute power. The power of the test is the probability of rejecting the
null hypothesis, assuming that the true population mean is equal to the
critical value. Since the region of acceptance is 294.45 to 305.55, the null
hypothesis will be rejected when the sampled run time is less than 294.45
or greater than 305.55.
Therefore, we need to compute the probability that the sampled run time will
be less than 294.45 or greater than 305.55. To do this, we make the following
assumptions:
The sampling distribution of the mean is normally distributed. (Because
the sample size is relative large, this assumption can be justified by the
central limit theorem.)
The mean of the sampling distribution is the critical value, 290.
The standard deviation of the sampling distribution is 2.83. The standard
deviation of the sampling distribution was computed in a previous lesson
Given these assumptions, we first assess the probability that the sample run
time will be less than 294.45. This is easy to do, using the Normal Calculator. We
enter the following values into the calculator: value = 294.45; mean = 290; and
standard deviation = 2.83. Given these inputs, we find that the cumulative
probability is 0.94207551. This means the probability that the sample mean will be
less than 294.45 is 0.942.
Next, we assess the probability that the sample mean is greater than 305.55.
Again, we use the Normal Calculator. We enter the following values into the
calculator: value = 305.55; mean = 290; and standard deviation = 2.83. Given these
inputs, we find that the probability that the sample mean is less than 305.55 (i.e.,
the cumulative probability) is 0.99999998. Thus, the probability that the sample
mean is greater than 305.55 is 1 - 0.99999998 or 0.00000002.
The power of the test is the sum of these probabilities: 0.94207551 +
0.00000002 = 0.94207553. This means that if the true average run time of the new
engine were 290 minutes, we would correctly reject the hypothesis that the run
time was 300 minutes 94.2 percent of the time. Hence, the probability of a Type II
error would be very small. Specifically, it would be 1 minus 0.942 or 0.058.
122
Test criteria
Criteria consisting of (i) specifying a level of significance ex, (ii) determining a
test statistic, (iii) determining the critical region(s), and (iv) determining the critical
value(s).
Test Statistic
The value of Z or t calculated for a sample statistic such as the sample mean
or the sample proportion.
Two tail test
A statistical hypothesis test in which the alternative hypothesis is stated in
such a way that it includes both the higher and the lower values of a parameter
than the value specified in the null hypothesis.
Type I error
An error caused by rejecting a null hypothesis that is true.
Type II error
An error caused by failing to reject a null hypothesis, that is not true.
9.3.9 Chi square test
Chi square is a non-parametric test of statistical significance for bivariate
tabular analysis (also known as cross breaks). Any appropriately performed test of
statistical significance lets you know the degree of confidence you can have in
accepting or rejecting a hypothesis. Typically, the hypothesis tested with chi square
is whether or not two different samples (of people, texts, whatever) are different
enough in some characteristic or aspect of their behavior that we can generalize
from our samples that the populations from which our samples are drawn are also
different in the behavior or characteristic.
A non-parametric test, like chi square, is a rough estimate of confidence; it
accepts weaker, less accurate data as input than parametric tests (like t-tests and
analysis of variance, for example) and therefore has less status in the pantheon of
statistical tests. Nonetheless, its limitations are also its strengths; because chi
square is more 'forgiving' in the data it will accept, it can be used in a wide variety
of research contexts.
Chi square is used most frequently to test the statistical significance of results
reported in bivariate tables, and interpreting bivariate tables is integral to
interpreting the results of a chi square test, so we'll take a look at bivariate tabular
(cross break) analysis.
Bivariate Tabular Analysis
Bivariate tabular (cross break) analysis is used when you are trying to
summarize the intersections of independent and dependent variables and
understand the relationship (if any) between those variables. For example, if we
wanted to know if there is any relationship between the biological sex of American
undergraduates at a particular university and their footwear preferences, we might
select 50 males and 50 females as randomly as possible, and ask them, "On
average, do you prefer to wear sandals, sneakers, leather shoes, boots, or
124
something else?" In this example, our independent variable is biological sex. (In
experimental research, the independent variable is actively manipulated by the
researcher; for example, whether or not a rat gets a food pellet when it pulls on a
striped bar. In most sociological research, the independent variable is not actively
manipulated in this way, but controlled by sampling for, e.g., males vs. females.)
Put another way, the independent variable is the quality or characteristic that you
hypothesize helps to predict or explain some other quality or characteristic (the
dependent variable). We control the independent variable (and as much else as
possible and natural) and elicit and measure the dependent variable to test our
hypothesis that there is some relationship between them. Bivariate tabular analysis
is good for asking the following kinds of questions:
1. Is there a relationship between any two variables in the data?
2. How strong is the relationship in the data?
3. What is the direction and shape of the relationship in the data?
4. Is the relationship due to some intervening variable(s) in the data?
To see any patterns or systematic relationship between biological sex of
undergraduates at University of X and reported footwear preferences, we co uld
summarize our results in a table like this:
Table: Male and Female Undergraduate Footwear Preferences
Sandals Sneakers Leather shoes Boots Other
Male
Female
Depending upon how our 50 male and 50 female subjects responded, we could
make a definitive claim about the (reported) footwear preferences of those 100
people.
In constructing bivariate tables, typically values on the independent variable
are arrayed on vertical axis, while values on the dependent variable are arrayed on
the horizontal axis. This allows us to read 'across' from hypothetically 'causal'
values on the independent variable to their 'effects', or values on the dependent
variable. How you arrange the values on each axis should be guided "iconically" by
your research question/hypothesis. For example, if values on an independent
variable were arranged from lowest to highest value on the variable and values on
the dependent variable were arranged left to right from lowest to highest, a positive
relationship would show up as a rising left to right line. (But remember, association
does not equal causation; an observed relationship between two variables is not
necessarily causal.)
Each intersection/cell--of a value on the independent variable and a value on
the independent variable--reports the result of how many times that combination of
values was chosen/observed in the sample being analyzed. (So you can see that
crosstabs are structurally most suitable for analyzing relationships between
125
nominal and ordinal variables. Interval and ratio variables will have to first be
grouped before they can "fit" into a bivariate table.) Each cell reports, essentially,
how many subjects/observations produced that combination of independent and
dependent variable values. So, for example, the top left cell of the table above
answers the question: "How many male undergraduates at University of X prefer
sandals?" (Answer: 6 out of the 50 sampled.)
Table 1.b. Male and Female Undergraduate Footwear Preferences
Sandals Sneakers Leather shoes Boots Other
Male 6 17 13 9 5
Female 13 5 7 16 9
Reporting and interpreting cross tabs is most easily done by converting raw
frequencies (in each cell) into percentages of each cell within the values/categories
of the independent variable. For example, in the Footwear Preferences table above,
total each row, then divide each cell by its row total, and multiply that fraction by
100.
Table 1.C. Male and Female Undergraduate Footwear Preferences (Percentages)
Sandals Sneakers Leather shoes Boots Other N
Male 12 34 26 18 10 50
Female 26 10 14 32 18 50
Percentages basically standardize cell frequencies as if there were 100
subjects/observations in each category of the independent variable. This is useful
for comparing across values on the independent variable, but that usefulness
comes at the price of a generalization --from the actual number of subjects/
observations in that column in your data to a hypothetical 100 subjects/
observations. If the raw row total was 93, then percentages do little violence to the
raw scores; but if the raw total is 9, then the generalization (on no statistical basis,
i.e., with no knowledge of sample-population representativeness) is drastic. So you
should provide that total N at the end of each row/independent variable category
(for replicability and to enable the reader to assess your interpretation of the table's
meaning).
With this caveat in mind, you can compare the patterns of distribution of
subjects/observations along the dependent variable between the values of the
independent variable: e.g., compare male and female undergraduate footwear
preference. (For some data, plotting the results on a line graph can also help you
interpret the results: i.e., whether there is a positive (/), negative (\), or curvilinear
(\/, /\) relationship between the variables.) Table 1.c shows that within our
sample, roughly twice as many females preferred sandals and boots as males; and
within our sample, about three times as many men preferred sneakers as women
and twice as many men preferred leather shoes. We might also infer from the
'Other' category that female students within our sample had a broader range of
footwear preferences than did male students.
126
The computation of chi square for each cell is listed below Table
Table Male and Female Undergraduate Footwear Preferences: Obse rved and Expected
Frequencies Plus Chi Square
Sandals Sneakers Leather shoes Boots Other Total
Male observed 6 17 13 9 5 50
Male expected 9.5 11 10 12.5 7
Female observed 13 5 7 16 9 50
Female expected 9.5 11 10 12.5 7
Total 19 22 20 25 14 100
Male/Sandals: ((6 - 9.5)2/9.5) = 1.289
Male/Sneakers: ((17 - 11)2/11) = 3.273
Male/Leather Shoes: ((13 - 10)2/10) = 0.900
Male/Boots: ((9 - 12.5)2/12.5) = 0.980
Male/Other: ((5 - 7)2 /7) = 0.571
Female/Sandals: ((13 - 9.5)2/9.5) = 1.289
Female/Sneakers: ((5 - 11)2/11) = 3.273
Female/Leather Shoes: ((7 - 10)2/10) = 0.900
Female/Boots: ((16 - 12.5)2 /12.5) = 0.980
Female/Other: ((9 - 7)2 /7) = 0.571
(Again, because of our balanced male/female sample, our row totals were the
same, so the male and female observed-expected frequency differences were
identical. This is usually not the case.)
The total chi square value for Table 1 is 14.026.
Interpreting the Chi Square Value
We now need some criterion or yardstick against which to measure the table 's
chi square value, to tell us whether or not it is significant. What we need to know is
the probability of getting a chi square value of a minimum given size even if our
variables are not related at all in the larger population from which our sample was
drawn. That is, we need to know how much larger than 0 (the absolute chi square
value of the null hypothesis) our table's chi square value must be before we can
confidently reject the null hypothesis. The probability we seek depends in part on
the degrees of freedom of the table from which our chi square value is derived.
Degrees of freedom
Mechanically, a table's degrees of freedom (df) can be expressed by the
following formula:
df = (r-1)(c-1)
132
That is, a table's degrees of freedom equals the number of rows in the table
minus one multiplied by the number of columns in the table minus one. (For 1 X 2
tables: df = k - 1, where k = number of values/categories on the variable.)
Degrees of freedom are an issue because of the way in which expected values
in each cell are computed from the row and column totals of each cell. All but one
of the expected values in a given row or column are free to vary (within the total
observed--and therefore expected) frequency of that row or column); once the free to
vary expected cells are specified, the last one is fixed by virtue of the fact that the
expected frequencies must add up to the observed row and column totals (from
which they are derived).
Another way to conceive of a table's degrees of freedom is to think of one row
and one column in the table as fixed, with the remaining cells free to vary. Consider
the following visuals (where X = fixed):
X X
(r-1)(c-1) = (3-1)(2-1) = 2 X 1 = 2
X X X
(r-1)(c-1) = (5-1)(3-1) = 4 X 2 = 8
So, for our Table 1, df = (2-1)(5-1) = 4
Sandals Sneakers Leather shoes Boots Other
Male X X X X X
Female X
column representing your previously chosen probability of error threshold (e.g., p <
.05) and across the row representing the degrees of freedom in your table. If your
chi square value is larger than the critical value in that cell, your data present a
statistically significant relationship between the variables in your table.
Table 1's chi square value of 14.026, with 4 degrees of freedom, handily clears
the related critical value of 9.49, so we can reject the null hypothesis and affirm the
claim that male and female undergraduates at University of X differ in their (self-
reported) footwear preferences.
Statistical significance does not help you to interpret the nature or explanation
of that relationship; that must be done by other means (in cluding bivariate tabular
analysis and qualitative analysis of the data). But a statistically significant chi
square value does denote the degree of confidence you may hold that relationship
between variables described in your results is systematic in the larger population
and not attributable to random error.
Statistical significance also does not ensure substantive significance. A large
enough sample may demonstrate a statistically significant relationship between two
variables, but that relationship may be a trivially weak one. Statistical significance
means only that the pattern of distribution and relationship between variables
which is found in the data from a sample can be confidently generalized to the
larger population from which the sample was ran domly drawn. By itself, it does not
ensure that the relationship is theoretically or practically important or even very
large.
Measures of Association
While the issue of theoretical or practical importance of a statistically
significant result cannot be quantified, the relative magnitude of a statistically
significant relationship can be measured. Chi -square allows you to make decisions
about whether there is a relationship netween two or more variables; if the null
hypothesis is rejected, we conclude that there is a statistically significant
relationship between the variables. But we frequently want a measure of the
strength of that relationship--an index of degree oof correlation, a measure of the
degree of association between the variables represented in our table (and data).
Luckily, several related measures of association can be derived from a table's chi
square value.
For tables larger than 2 X 2 (like our Table 1), a measure called 'Cramer's phi'
is derived by the following formula (where N = the total number of observations, and
k = the smaller of the number of rows or columns):
Cramer's phi = the square root of (chi-square divided by (N times (k minus 1)))
So, for our Table 1 (a 2 X 5), we would compute Cramer's phi as follows:
1. N(k - 1) = 100 (2-1) = 100
2. chi square/100 = 14.026/100 = 0.14
3. square root of (2) = 0.37
134
9.6 SUMMARY
This chapter had given clear picture about the process of Hypothesis Testing
and to find out the details of Null hypothesis and alternate hypothesis. This lesson
also has given the significance and differences between Type I error and Type II
error. The detailed step-wise procedures were given to perform the hypothesis
testing. This chapter also gave details about the calculation of P value. The above
chapter has also given the framework for performing key statistical tests like chi -
square test. Chi-square is a non parametric test of statistical significance for
bivariate tabular analysis
9.7 TERMINAL EXERCISES
1. -----------------is an assumption about a population parameter.
2. ------------------ is usually the hypothesis that sample observations result
purely from chance effects.
3. ----------- is the hypothesis that sample observations are influenced by some
non-random cause.
4. The --------------- is rejected if p is at or below the significance level; it is not
rejected if p is above the significance level.
5. A true null hypothesis can be incorrectly rejected is known as ------- error
6. A false null hypothesis can fail to be rejected is known as ------- error
9.8 SUPPLEMENTARY MATERIALS
1. Cronin, M. J. (1998a, Feb. 2). Business secrets of the billion -dollar website.
Fortune, p. 142.
2. Cohen, J. (1990). Things I have learned (so far). American Psychologist,
1304–1312
3. Chicago Manual of Style, 14th ed. (1993). Chicago: University of Chicago
Press.
9.9 ASSIGNMENTS
1. What is a hypothesis? What characteristics it must possess in order to be a
good research hypothesis?
2. A manufacturer considers his production process to be working properly if
the mean length of the rods the manufactures is 8.5". The standard
deviation of the rods always runs about 0.26". Suppose a sample of 64 rods
is taken and this gives a mean length of rods equal to 8.6". What are the null
and alternative hypotheses for this problem? Can you infer at 5% level of
significance that the process is working properly?
3. The procedure of testing hypothesis requires a researcher to adopt several
steps. Describe in brief all such steps.
4. Point out the important limitations of tests of hypotheses. What precaution
the researcher must take while drawing inferences as per the results of the
said tests?
136
LESSON – 10
H1 :
Assumptions for the Test to Determine the Difference between Two Means
The samples must be independent of each other. That is, there can be no
relationship between the subjects in each sample.
The populations from which the samples come must be (approximately)
normally distributed or the sample sizes of both groups should be at least
30.
The standard deviations of both populations must be known.
We can compare the groups by the difference in their population means , µ1 -
µ2, where µ1 is the population mean for group 1 and µ2 is the population mean for
group 2.
We estimate µ1-µ2 with
139
When both populations are normally distributed or the samples size for each
group is at least 30, then
The observed difference between the sample means may be due to chance, in
which case the null hypothesis will not be rejected. If the difference is statistically
significant, the null hypothesis is rejected an d the researcher can conclude the
population means are different. The common approach to finding critical values
and P-values that was used here .
Example: Dr. Cribari would like to determine if there is a statistically
significant difference between her two Math 2830 classes. To make this comparison
she will compare the results from exam 1. Class one had 35 students take the exam
with a mean of 82.6 and a population standard deviation of 1.41. Class two had 32
students take the exam with a mean of 84 and a population standard deviation of
3.63. Can Dr. Cribari conclude that there is difference in the mean test grades
between the two classes at ∞=0.05?
Step 1: State the hypotheses and identify the claim
Step 2: Find the critical value(s) from the appropriate table .As stated, the
problem is giving the population standard deviations. This means that we will be
doing a z-test.
Two sided test => critical value = ±1.96 ∞ 0.05
Step 3: Compute the test value and determine the P-value.
140
Step 4: Make the decision to reject or not reject the null hypothesis.
Since the p-value is smaller than our ∞, the null hypothesis is rejected.
[OR, Since, our test value, -2.05, falls within the rejection region, the null
hypothesis is rejected]
Step 5: Summarize the results.
That is, there is evidence to support the claim that the exam 1 grades differ
between the two sections of MATH2830.
Practice Example: A survey found that the average hotel room in New Orleans
is $88.42 and the average room rate in Phoenix is $80.61. Assume that the data
were obtained from two samples of 50 hotels each and that the (population)
standard deviations were $5.62 and $4.83, respectively. At ∞ = 0.01, can it be
concluded that the average hotel room in New Orleans costs more than in Phoenix?
Step 1: State the hypotheses and identify the claim.
Step 2: Find the critical value(s) from the appropriate table.
Step 3: Compute the test value and determine the P-value.
Step 4: Make the decision to reject or not reject the null hypothesis.
Step 5: Summarize the results
Formula for the z Confidence Interval for Difference Between Two Means
Assumptions
The data for each group are independent random sample s.
The data are from normally distributed populations and/or the sample
sizes of the groups are greater than 30.
The population standard deviation is (assumed) known
Example: Two brands of cigarettes are selected, and their nicotine content is
compared. The data are shown below. Find the 95% confidence interval of the true
difference in the means.
141
Warning: Your calculator will perform a 2 sample t-test (its #4 under STATS
then TESTS). However, it uses a complicated formula to determine the degrees of
freedom that will ultimately affect how the calculator deals with confidence
intervals and p-values. We will come back to this point at the end of the section.
Example: A real estate agent wishes to determine whether tax assessors and
real estate appraisers agree on the values of homes. A random sample of the two
groups appraised 10 homes. Is there a significant difference in the values of the
142
homes for each group? Let ∞ = 0.05. Assume the data are from normally distributed
populations.
Step 4: Make the decision to reject or not reject the null hypothesis.
The null hypothesis is rejected. This decision can be based on:
The fact that the test value (-4.02) is within the critical region since it is
less than -2.262 or
The fact that the p-value (0.003) is smaller than a=0.05
143
Warning: The way our calculator determines the degrees of freedom is not the
same as the book. So you will NOT be able to use your calculator STAT/TESTS
function to calculate your confidence interval because you will get a VERY different
confidence interval. This is due to the fact that the t-multiplier will be sufficiently
different then what the calculator will find.
Example: Let’s find the 95% Confidence Interval for the first problem.
Example: Let’s find the 99% Confidence Interval for the second problem.
10.4 Testing the Difference Between Two Means: Dependent Samples
So far we have only compared two means when the samples were independent.
Samples are considered to be dependent when the subjects are paired or matched
in some way
Examples of paired data
Each person is measured twice where the 2 measurements measure the
same thing but under different conditions
Similar individuals are paired prior to an experiment and each member of a
pair receives a different treatment
Two different variables are measured for each individual and there is
interest in the amount of difference between the 2 variables
When using paired data, you are interested primarily in the "difference" and not the
data itself
When samples are dependent, a special t test for dependent means is used.
The test uses the difference in the values of the matched pairs.
Important: We cannot use the t test we had learned for a difference in
independent means.
To determine whether one set of observations tend to be larger or different
than the paired observations, we take the difference between the matched
observations and perform analysis on the differences.
Classic example would be weight loss
Weight before
Weight after
We are interested in change
An aside: this study also used a placebo group. Why?
145
Here, µ1 is the mean of the population of the first set of measurements and µ2
is the mean of the population of the second set of measurements.
Actually, you can also use µd = µ2 -µ1 long as you are consistent with your
statement of hypotheses and calculation of D.
Formulas for the t Test for Dependent Samples
146
The good news is we can find the mean of the differences, ¯D, and the
standard deviation of the differences, sD , using the LIST and STAT functions in
your TI-83/84.
1. Go to STAT -> EDIT -> Edit .
2. Enter the first set of observations under L1.
3. Enter the second set of observations under L2.
4. Highlight L3 in list, type L1 - L2 and hit enter. The set of differences should
now be calculated.
5. Go to STAT -> CALC -> 1-Var Stats, hit enter. Type L3 (after 1-Var Stats on
your screen) and hit enter. Your calculator will calculate the sample mean
and standard deviation for you .
Example: A physical education director claims by taking a special vitamin, a
weight lifter can increase his strength. Eight athletes are selected and given a test
of strength, using the standard bench press. After 2 weeks of regular training,
supplemented with the vitamin, they are tested again. Test the effectiveness of the
vitamin regimen at a = 0.05. Each value in these data represents the maximum
number of pounds the athlete can bench press. Assume the variable is
approximately normally distributed.
Before X1 After
210 219
230 236
182 179
205 204
262 270
253 250
219 222
216 216
Step 1: State the hypotheses and identify the claim.
I will base my differences on: D = strength after - strength before.
Ho: µD=0 H1: µD>0
Step 2: Find the critical value(s) from the appropriate table.
We have 8 lifters which gives 7 degrees of freedom. Our ∞=0.05.
We have a right tailed test => critical value will be positive. t critical value =
1.895
147
Step 4: Make the decision to reject or not reject the null hypothesis
The null hypothesis is NOT rejected. We can base this decision on either of the
two facts:
The p-value is larger than ∞ = 0.05
The test value (1.388) is smaller than the critical value (1.895). That is, our
test value is within the non-rejection region.
Step 5: Summarize the results
There is not sufficient evidence to support the education director claims by
taking a special vitamin, a weight lifter can increase his strength.
Practice Example: A sample of 10 college students in a class were asked how
many hours per week they watch TV and how many hours a week they used a
computer. Is there a difference in the mean number of hours a college student
148
Student Comp TV
1 30 2
2 20 1.5
3 10 14
4 10 2
5 10 6
6 0 20
7 35 14
8 20 1
9 2 14
10 5 10
10.6 SUMMARY
Basic data analysis provides valuable insights and guides the rest of the data
analysis as well as interpretation of results. The parametric and non parametric
tests are also available for testing the hypothesis related to difference. In parametric
case, the t and z test is used to determine hypothesis related to population mean.
Different forms of t and z test are suitable for testing hypothesis based on sample
and two independent samples or paired samples.
10.7 TERMINAL EXERCISES
1. The observed difference between the sample means may be due to ------
2. A----------test is used to test the difference between means when the two
samples are independent and when the samples are taken from two
normally or approximately normally distributed populations
3. Each person is measured twice where the 2 measurements measure the
same thing but under different conditions is known as ---------
4. We cannot use the ----- test we had learned for a difference in independent
means
10.8 SUPPLEMENTARY M ATERIALS
a) Identify the most appropriate statistical test
b) Identify the independent variable(s) or factor(s)
c) Identify the dependent variable
d) Identify the scale of measurement for the independent and dependent
variables.
10.9 ASSIGNMENTS
Research suggests that the antioxidants in foods such as blueberries can
reduce perhaps reverse age-related declines in cognitive functioning. To test this
phenomenon, a researcher selects a sample of 25 adults aged between 70 and 75
years old and administers a cognitive function test to each participant. The
participants then drink a blueberry supplement every day for 4 months before they
are tested again. The researcher compares the scores before treatment with the
scores after treatment to see if there is any change in cognitive function.
10.10 SUGGESTED READING/REFERENCE BOOKS
1. De Vaus, D.A. (1991) Surveys in Social Research, 3rd edition, London: UCL
Press.
2. Gash, S. (1999) Effective Literature Searching for Research, 2nd edition,
Aldershot: Gower.
3. Hart, C. (2001) Doing a Literature Search, London: Sage.
Spence, G. (2001) A Simple Guide to Internet Research, Harlow: Prentice Hall
150
LESSON – 11
F- TEST (ANOVA)
11.1 INTRODUCTION
Analysis of variance (abbreviated as ANOVA) is an extremely useful technique
concerning researches in the fields of economics, biology, education, psychology,
sociology, business/industry and in researches of several other disciplines. This
technique is used when multiple sample cases are involved. As stated earlier, the
significance of the difference be tween the means of two samples can be judged
through either z-test or the t-test, but the difficulty arises when we happen to
examine the significance of the difference amongst more than two sample means at
the same time. The ANOVA technique enables us to perform this simultaneous test
and as such is considered to be an important tool of analysis in the hands of a
researcher. Using this technique, one can draw inferences about whether the
samples have been drawn from populations having the same mean.
The ANOVA technique is important in the context of all those situations where
we want to compare more than two populations such as in comparing the yield of
crop from several varieties of seeds, the gasoline mileage of four automobiles, the
smoking habits of five groups of university students and so on. In such
circumstances one generally does not want to consider all possible combinations of
two populations at a time for that would require a great number of tests before we
would be able to arrive at a decision. This would also consume lot of time and
money, and even then certain relationships may be left unidentified (particularly
the interaction effects). Therefore, one quite often utilizes the ANOVA technique and
through it investigates the differences among the means of all the populations
simultaneously.
11.2 OBJECTIVES
After studying this lesson, you will be able to understand:
ANOVA
Importance of ANOVA
Assumptions
Conditions
ANOVA model
10.3 CONTENTS
10.3.1 What is ANOVA?
10.3.2 Importance of ANOVA
10.3.3 Assumptions
10.3.4 Conditions for ANOVA
10.3.5 ANOVA Model
152
Some experiments have more than one between -subjects factor. For instance,
consider a hypothetical experiment in which two age groups (8 -year olds and 12-
year olds) are asked to perform a task either with or without distracting
background noise. The two factors are age and distraction.
11.3.2 Importance of ANOVA
It is important to know about analysis of variance because it is the most
common inferential statistical procedure used in experiments. Why? Because there
are actually many versions of ANOVA, so it can be used with many different
designs: It can be applied to an experiment involving independent samples or
related samples, to an independent variable involving any number of conditions,
and to a study involving any number of independent variables. Such complex
designs are common because, first, the hypotheses of the study may require
comparing more than two conditions of an independent variable. Second,
researchers often add more conditions because, after all of the time and effort
involved in creating two conditions, little more is needed to test additional
conditions. Then we learn even more about a behavior (which is the purpose of
research). Therefore, you’ll often encounter the ANOVA when conducting your own
research or when reading about the research of others.
11.3.3 Assumptions
Analysis of variance assumes normal distributions and homogeneity of
variance. Therefore, in a one-factor ANOVA, it is assumed that each of the
populations is normally distributed with the same variance (σ²). In betwe en-
subjects analyses, it is assumed that each score is sampled randomly and
independently. Research has shown that ANOVA is "robust" to violations of its
assumptions.
This means that the probability values computed in an ANOVA are
satisfactorily accurate even if the assumptions are violated. Moreover, ANOVA tends
to be conservative when its assumptions are violated. This means that although
power is decreased, the probability of a Type I error is as low or lower than it would
be if its assumptions were met. There are exceptions to this rule. For example, a
combination of unequal sample sizes and a violation of the assumption of
homogeneity of variance can lead to an inflated Type I error rate.
11.3.4 Conditions for ANOVA
1. The sample must be randomly selected from normal populations
2. The populations should have equal variances
3. The distance from one value to its group’s mean should be independent of
the distances of other values to that mean (independence of error).
4. Minor variations from normality and equal variances are tolerable.
Nevertheless, the analyst should check the assumptions with the diagnostic
techniques.
154
than 1. The f distribution determines the size of ratio necessary to reject the null
hypothesis for a particular sample size and level of significance.
11.3.5 ANOVA model
To illustrate reports one way ANOVA, consider the following hypothetical
example. To find out the number one best business school in India, 20 business
magnets were randomly selected and asked to rate the top 3 B-schools. The ratings
are given below
Data
Table on means
evidence to show that the true porportion is lower than 60% ? Conduct the test at
10% Type I error rate, and use the p-value and rejection region approaches.
11.10 LEARNING ACTIVITIES
1. Which procedure is more useful in marketing research –ANOVA or Chi
Square test or T – test? Discuss as a small group
11.11 SUGGESTED READINGS/ REFERENCE BOOKS
1. Campbell, A. A., & Katona, G. (1966). The sample survey: A technique for
social science research. In L. Festinger & D. Katz (Eds.), Research methods
in the behavioral sciences. New York: Holt, Rinehart and Winston.
2. Cronbach, L. J. (1990). Essentials of psychological testing (5th ed.).
New York: Harper & Row.
3. Davies, G. R., & Yoder, D. (1937). Business statistics. New York: John Wiley.
4. Davis, D., & Cosenza, R. M. (1988). Business research for decision
making (2nd ed.). Boston: PWS-Kent Publishing.
11.12 KEYWORDS
One-tailed test: The test used to evaluate a statistical hypothesis that predicts
that scores will only increase or only decrease
One-way ANOVA: The analysis of variance performed when an experiment has
only one independent variable
Nonsignificant: Describes results that are considered likely to result from
chance sampling error when the predicted relationship does not exist; it indicates
failure to reject the null hypothesis
Probability: A mathematical statement indicating the likelihood that an event
will occur when a particular population is randomly sampled; symbolized by p
Mean square: In ANOVA, an estimated population variance, symbolized by MS
F-distribution: The sampling distribution of all possible values of F that occur
when the null hypothesis is true and all conditions represent one population
Fisher’s protected t-test: The post hoc procedure performed with ANOVA to
compare means from a factor in which all levels do not have equal ns
F-ratio: In ANOVA, the ratio of the mean square between groups to the mean
square within groups
160
LESSON – 12
Scatter Plot
The first step is creating a scatter plot of the data. “There is no excuse for
failing to plot and look.” In general, scatter plots may reveal a
• Positive Correlation (high values of X associated with high values of Y)
• Negative Correlation (high values of X associated with low values of Y)
• No Correlation (values of X are not at all predictive of values of Y).
These patterns are demonstrated in the figure to the right.
Steps for Calculation of p
1. First of all it is required to assign position of merit or rank to each
individual joins either test. These ranks are put under column 3. (Designated as R1 )
and 4 (designated as R2 ) respectively. The task of assigning ranks in the cases like
example 1st is not difficult. But in the cases, like example 2nd, where two or more
individuals are found to achieve same score, some difficulty arises. In the above
example, in the first test X, B and F are two individuals who have the same score
i.e. 15. Therefore, score 15 occupies 6th position in order of merit. But now the
question arises which one of the two individuals B and F should be ranked as 6th
or 7th. In order to overcome this difficulty we equally share the rank 5th and 7th
between them and thus rank each one of them as 6.5.
165
1. Similarly, if there are three persons who have the same score and share the
same ranks, we take the average of the ranks claimed by these persons. For
example, we can take the score 20 in the 2nd example which is shared by three
individuals D, G and H. It is ranked third in the whole series and therefore the
ranks 3, 4 and 5 are shared equally by D, G and H and hence we attribute rank to
each of them.
2. After writing down the allotted rank to all the individuals 0 either of the two
tests, the differences in these ranks are calculated. In doing so we do not 'consider
the algebraic signs +ve or -ve of the difference. This difference is written under
column 5th (designated as d 2)
3 In the next column (designated as d 2) we square up the Rank difference or
the values of d written in the column 5th.
4. Now the value of p is calculated by the formula, where d2 stands for the sum
of the squares of differences between the ranks of the scores on two different tests
and N for the number of individuals whose scores are under consideration for
computing.
Where x and y represent the deviation of scores in the tests X and Y from the
means of each distribution. The procedure of calculation r by this formula can be
understood by the following illustration
Individuals Scores in Scores in X Y Xy X2 2
Y
Computation of r directly from raw scores when deviations are take n from zero
(without calculating deviations from the means. Here we apply the formula.
Scores in 1st Scores in 2nd
Subject
Test (X) Test (Y) XY X 2 Y 2
A 5 12 60 25 144
B 3 15 45 2 225
C 2 11 22 4 121
D 8 10 80 64 100
E 6 18 108 36 324
quantities, and we could label them as: input and output, independent and
dependent, or explanatory and response. For example, the family income is the
independent or explanatory variable while the food expenditure will be called the
dependent or response variable.
The response variable is that variable whose value can be explained by the
value of one or more explanatory variable or predictor variables. There are many
types of regression, based on the number of variables involved. When we have one
response and one predictor, this is the case for linear regression. In case we have
one response and many predictor variables, this is the case of multiple regressions.
In this chapter, we are interested in the case of two variables, and when the
relationship between them is assumed to be a linear one. In addition to linear
regression between two variables there might be logarithmic, exponential,
quadratic, cubic, and so on regression.
The purpose of regression analysis is to determine the existence, the kind, or
the extent of a relationship (in the form of a mathematical equation) between two or
more variables. For example, if x and y denote two variables under study, then, in
general, we want to determine the best equation (relation) y = f(x) that describes the
relationship between x and y.
The term "regression" goes back to Sir Francis Galton. In his work on
hereditary, Galton noted that the sons of very tall fathers tended to be tall, but n ot
quite as tall as their fathers. Also, sons of very short fathers tended to be short, but
not quite as short as their fathers. He called this tendency, of individuals to be not
quite tall or as short as their fathers, the tendency of regression toward mediocrity",
i.e., going back to the average.
In much of the experimental work, it is desired to investigate how the changes
in one variable affect another variable, or introduce some changes in that variable.
One can distinguish between exact and non -exact relationships.
Exact Relationship
Sometimes we find that two variables are linked by an exact relationship. For
example, if the resistance "R" of a simple circuit is kept constant, the current "I" is
related to the voltage "V" by Ohm's Law: I = V/R, which is an equation of first
degree and the graph of it, is a straight line. This is an exact law. However, if we
want to verify it empirically by making changes in V and observing I while R is kept
constant we notice that the plot of the empirical points (V, I) will not fall exactly on
a straight line. The reason is that the measurements maybe subject to slight errors
and thus the plotted points would not fall exactly on the straight line but would
vary randomly around it. For purposes of predicting the value for I for a particular
value for V (with R as a fixed constant) we should use the plotted straight line .
Non-Exact Relationship
Sometimes the relationship is not exact even apart from the errors in
measurements. For example, suppose we consider the height and weight of adult
168
males for some population. If we plot the ordered pair (weight, height) = (x, y), a
diagram, like Figure 1 below, will result. Such a representation is conventionally
called a scatter diagram
Note that for any given height there is a range of observed weights and vice
versa. This variation will be partially due to measurement errors but primarily due
to variation among individuals. If you consider two males with same height, their
weights will be different not only because we cannot me asure the weights exactly
correct, but also because the two males will have slight different weights. Thus no
unique relationship between the actual weight and height can be written. But we
can notice that the average observed weight, for a given observed height,
increases as height increases. Whether a relationship is exact or non -exact, in so
far as average values are concerned, it will be useful especially for prediction
purposes.
It is not always clear which variable should be considered the response
variable and which the explanatory variable. For example, does high school GPA
predict a student's SAT score or can the SAT score be used to predict the student's
GPA? The investigator should determine which variable plays the role of the
explanatory variable based on the question that needs to be answered.
Scatter Diagrams show the type of the relation that exists between two
variables. By plotting the explanatory variable along the horizontal axis and the
response along the vertical axis, our goal is to distinguish which scatter diagrams
will imply a linear relation from those that will imply a nonlinear relation and those
that imply no relation. Figure 2 and Figure 3 show some scatter diagrams and the
type of relation implied.
169
Figure – 4
Figure 4, above, shows the type of linear and nonlinear relationships between
two variables.
In Figure 4A we see a perfect negative linear relationship, and we say that the
two variables are negatively associated. In other words, two variables are negatively
associated if, whenever the value of one variable increases, the value of the other
variable decreases.
On the other hand, Figure 4B shows that there are two variables that are
linearly related and positively associated. In the same vein, two variables are
170
positively associated if, whenever the value of one variable increases, the value of
the other variable increases.
The situation is quite different in Figure 4C. There is the case where we see a
horizontal line, although it is linear, but the response variable, Y, is not affected by
the change in the explanatory variable, x. Thus Y = a constant.
Figure 4D shows a non-linear relationship between the two variables x and y.
Figure 4 displays ideal cases for positively and negatively associated variables
based on a linear relationship between the two. Later we will check on the strength
of that linear relationship between the explanatory variable x and the response
variable Y.
2.7.1 Regression Models
The simplest linear regression model is of the form
Figure 5
171
x i
x i 1
n
And
n
y i
y i 1
n
Each observation (xi, y i), i = 1, 2,..., n, in the sample, satisfies the equation
Yi = 0 + 1x1+i
Where i is the value assumed by Ei when Yi takes on the value y i. The above
equation can be viewed as the model for a single observation y. Similarly using the
estimated, or fitted regression line
Where e i = y i - y1 is called a residual and describes the error in the fir of the
model at the ith data point or observation
We shall find b0 and b1, the estimates of 0 and 1, so that the sum of the
squares of the residuals is a minimum. The residual sum of squares is often called
the sum of squares of the errors about the regression line and denoted by SSE. This
minimization procedure for estimating the parameters is called the Method of Least-
Squares. Thus we shall find b0 and bl so as to minimize
172
n n n
SSE e 2 i ( yi yˆ1 ) 2 ( yi b0 b1 xi ) 2
i 1 i 1 i 1
(The interested reader, who is aware of differential calculus, can check the
procedure in R.E. Walpole, and R.H. Myers; Probability and Statistics for Engineers
and Scientists, 4 th Edition, p. 361.) The least -squares regression line is the line
that minimizes the square of the vertical distance between the observed values of Y
and those predicted by the line, ŷ Thus solving the normal equations resulting from
the differentiation of the SSE, we see that
n n
n ( xi )( yi )
x , y i 1
i 1
i 1
i i
n
b1
n
2
x
n 2 i 1 i
x i
i 1
, and b0 = yˆ b1x
EXAMPLE 5.1
The following measurements of the specific heat of a certain chemical were
made in order to investigate the variation in specific hear with temperature.
Tempo 0 10 20 30 40
X
il
i Yi = 59.8. Thus by applying the above formulas for bo b1 , we have bo 0.514 ,
and b1 0.0028. Hence the fitted equation will be given by y = 0.514 0.0028x.
When temperature x 25 C the predicted specific heat is 0.514-00028*25 = 0.584.
0
n n
bi xi yi / x 2i , andb0 y
i 1 i 1
If there are an odd number of evenly spaced data values for X, denote them by:
-3, -2, -1, 0, 1, 2, 3 ...
If there is an even number of evenly spaced data values for X, denote them by:
-5, -3, -1, 1,3, 5, ....
Consider the following data that stands for the output of a certain company for
the years 1960-1964. Predict the output of the company in the year 1965.
Solution
Based on the above setup we can see that the data can be coded to look like
the following table.
Output (1000 tons) 11.1 12.3 13.7 14.6 15.6
Year -2 -1 0 1 2
For manual calculations, we have the following Table.
x2 4 1 0 1 4 10
Table
Hence the estimated values in the predicted equation will have the following
values: b0 = 13.46 and b =1.13 and the fitted equation is given by ŷ = 13.46 +
1.13x, where x is the coded year.
To predict the output for the year 1965, we take x = 3, which is its coded
Where P is the population size at time t, A is the population size at time zero (a
constant), "r" is a constant representing the rate (of growth when r>0 and decay
when r<0) and u represents a random error. This model describes situations when
the relative growth is constant in the population. It can describe the growth in
human or biological populations, and the growth in compound interest, when u =1.
Consider the following data for the population of the USA (in millions) during
the years 1860-1900:
Year 1860 1870 1880 1890 1900
Population 31.4 39.8 50.2 63.0 76.0
Which can be written as: y = 0 + 1x + E, where y = Ln P, 0 = r, x = t and E =
Ln u = 0. Now we have a linear mode l, we can fit using the data given above.
Moreover, by appealing to EXAMPLE 5.2, and coding the data, we have the
following generated data in Table 2 for manual calculations.
Equivalently we have
ŷ
Yi - y = ( i - y ) + (Yi -
ŷi ), for i = 1,2, 3, ….., n.
In other words, we can write
Total deviation = Deviation due to regression + Deviation about regression
Squaring both sides and summing over i = 1, 2, ....n, we have (interested
reader might need to check this identity by recalling the identity
(x
i 1
1 x 0)
n n n
n n Y 2
i
SStotal (Yi y ) Y i 2 2 i 1
i 1 i 1 n
n
n n n Y 2
i
i 1 i 1 i 1
176
What is that r2? It is not surprising to see that it is the square of the linear
correlation coefficient found above, when it needs to be calculated. If r 2 is found
first, clearly we can find r by taking the square root. Another que stion: is r > 0 or r
< 0? How do you decide? No doubt that finding r first is more enlightening than
finding r2 first. Here is another definition.
The coefficient of determination, r2 , measures the percentage of total variation
in the response variable that is explained by the least square regression line based
on the one explanatory variable x. it is a number between 0 and 1 inclusive, that is,
0 <r2 < 1. Recall that r2 is the estimated value of the random variable R 2. Practically
r2 is expressed as a percentage, and it is < 100% . What about that difference
percentage between the 100% and r 2? This difference is the unexplained variation
in Y due to not including all the variables that the response variable can get
affected by.
The results are summarized in the ANOVA Table below:
Source df SS Ms E(MS)
Regression 1 SSregr n
1
SSregr - b 2 X X)2
1 i
i 1
Residual n-2 SSresidual SStotal - SSregr SSresidual
-2
Total n-2 SStotal (See Bakir and Shayib
1990)
Testing for Significant Regression
In the model y = 0 + 1x + e, if 1 = 0, then Y is not related to x by the first
linear model, i.e., this first order model is incorrect. However if β1 ≠ 0, then there is
a linear relationship between x and y as described in the model. At this time we
cannot tell whether the relationship is positively related, i.e. y increases as x
increases, or negatively related, i.e., x increases then Y decreases. Thus to test for
significant linear (first order) relationship, between x and Y is equivalent to testing
the following hypotheses:
SS
regr
MSregr , and
1
SS
MSresidual residuals
-2
177
n (Xi - X) 2
i 1
2
E (b1 ) 1 , Var (b1 ) n
n (Xi - X)2
i 1
n
X1
2
(b o ) t1 - /2 MSresiduals n i 1
2
n (X i - X)
i 1
And the 100 (1 - ) % on 1 is:
b1 t1 - /2 n
MSresiduals
2
n (X i - X)
i 1
When 1 /2 t −α is read from the t-table with n–2 degrees of freedom = residual
degrees of freedom.
Remember there are equivalent forms for the above intervals in terms of the
standard error.
178
For more C.I. on the mean of the response variable at a given value for the
explanatory variable, or for a C.I. for the regression line at a given x value or for a
C.I. on a predicted single y value at a specified x value, we refer the interested
reader to Bakir and Shayib 1990, pp. 172–176. In addition to the above the
interested reader can check that the random variable T2 has an F-distribution.
From the model: y = β0 + β1x + e, again for the sake of completion, with the
usual assumptions for the E’s that are independent random variables with mean
zero and variance σ2, let us give, without derivation, the 100(1 –α) % C.I. for a mean
response, i.e. on E(Y) = β0 + β1x, which depends on the x-value when it is given.
The following formula can be used to construct a confidence interval on μ y|x
When the specified value for the explanatory variable is x0, n is the sample
size, and 1 /2 t −α is the critical value, with n–2 degrees of freedom.
The procedure for obtaining a prediction interval for an individual response is
identical to that for finding a C.I. on the mean of that response. The difference is in
the standard error. More variability is associated with one value than with the
mean of some values. he following formula can be used to construct a prediction
interval for a predicted y value at a specified x-value, x = x0.
12.4 REVISION POINTS
1. The size of r is very much dependent upon the variability of measured values
in the correlated sample.
2. The relationship between two sets of scores or variables can be represented
graphically by a straight line, it is known as Linear Correlation.
3. Correlation coefficients assume the relation can be described with a straight
line.
4. Coefficient Of Correlation is a kind of ratio which expresses the extent to
which changes in one variable are accompanied with changes in the other
variable. It involves no units and varies from -1 (indicating perfect negative
correlation) to + 1 (indicating perfect positive correlation). In case the
coefficient of correlation is zero it indicates zero correlation between two sets
of measures.
5. The response variable is that variable whose value can be explained by the
value of one or more explanatory variable or predictor variables.
179
181
UNIT – IV
LESSON – 13
3. Interval scale: This scale also uses numbers such that these can be ordered
and their differences have a meaningful interpretation.
4. Ratio scale: A scale possessing all the properties of an interval scale along
with a true zero point is called a ratio scale. It may be pointed out that a zero point
in an interval scale is arbitrary. For example, freezing point of water is defined at
0° Celsius or 32° Fahrenheit, implying thereby that the zero on either scale is
arbitrary and doesn’t represent total absence of heat. In contrast to this, the
measurement of distance, say in metres, is done on a ratio scale. The term ratio is
used here because ratio comparisons are meaningful. For example, 100 kms of
distance is four times larger than a distance of 25 kms while 100°F may not mean
that it is twice as hot as 50°F. It should be noted here that a test that can be
performed on high level measurements can always be performed on ordinal or
nominal measurements but not vice-versa. However, if along with the high level
measurements the conditions of a parametric test are also met, the parametric test
should invariably be used because this test is most powerful in the given
circumstances. From the above, we conclude that a non -parametric test should be
used when either the conditions about the parent population are not met or the
level of measurements is inadequate for a parametric test.
A parametric statistical test is one that makes assumptions about the
parameters (defining properties) of the population distribution(s) from which one's
data are drawn.
A non-parametric test is one that makes no such assumptions. In this strict
sense, "non-parametric" is essentially a null category, since virtually all statistical
tests assume one thing or another about the properties of the source popu lation(s).
Which is more powerful?
Non-parametric statistical procedures are less powerful because they use less
information in their calculation. For example, a parametric correlation uses
information about the mean and deviation from the mean while a n on-parametric
correlation will use only the ordinal position of pairs of scores.
Parametric Assumptions
The observations must be independent
The observations must be drawn from normally distributed populations
These populations must have the same variances
The means of these normal and homoscedastic populations must be linear
combinations of effects due to columns and/or rows
Nonparametric Assumptions
Certain assumptions are associated with most nonparametric statistical tests,
but these are fewer and weaker than those of parametric tests.
183
In research activities a YES/NO scale is nominal. It has no order and there
is no distance between YES and NO.
2. Ordinal or Ranking Scale
There is no objective distance between any two points on your subjective
scale.
3. Interval Scale
It is an
interval scale because it is assumed to have equidistant points between each of the
scale elements.
4. Ratio Scale
Ratio data is interval data with a natural zero point
13.4 REVISION POINTS
1. Level of measurement
2. Parametric test
3. Non parametric test
4 Advantage and disadvantage of non parametric test
13.5 INTEXT QUESTIONS
1. List out the advantages and disadvantages of non parametric test
2. Explain the limitations of nonparametric tests
13.6 SUMMARY
After going through this Lesson you should be able to explain Level of
measurement, Parametric test, Non parametric test, Advantage and disadvantage of
non parametric test
13.7 TERMINAL EXERSICE
1. Compare and contrast parametric and nonparame tric tests
2. Identify multiple applications where nonparametric approaches are
appropriate
185
LESSON – 14
SIGN TEST
14.1 INTRODUCTION
One of the easiest non-parametric tests is the sign test. The test is known as
the sign test as it is based on the direction of the plus or minus signs of
observations in a sample instead of their numerical values. There are two types of
sign tests: (a) One-sample sign test,and (b) Two-sample sign test.
14.2 OBJECTIVE
The main aim of this Lesson is to study the Non parametric test. After going
through this Lesson you should be able to understand Sign test, one sample sign
test and two sample sign test
14.3 CONTENT
14.3.1 Sign test
14.3.2 One sample sign test
14.3.3 Two sample sign test
14.3.1 one -sample sign test
The one-sample sign test is a very simple non -parametric test applicable on
the assumption that we are dealing with a population having a continuous
symmetrical distribution. As such, the probability of getting a value less than the
mean is 0.5. Likewise, the probability of getting a value greater than the mean
isalso 0.5. To test the null hypothesis µ = µ0against an appropriate alternative,
each sample value greater than µ0is replaced by plus (+) sign and each sample
value less than µ0with a minus (-) sign. Having done this, we can test the null
hypothesis that the probabilities of getting both plus and mi nus signs are 0.5. It
may be noted that if a sample value happens to be equal to µ0, it is simply
discarded.
To perform the actual test, we use either of the two methods. When the sample
is small, the test is performed by computing the binomial probabilities or by
referring to the binomial probabilities table. When the sample is large, the normal
distribution is used as an approximation of the binomial distribution. Let us take
an example to show how the onesample sign test is applied.
Example 1
We are required to test the hypothesis that the mean value µ of a continuous
distribution is 20 against the alternative hypothesis µ ≠20. Fifteen observations
were taken and the following results were obtained:
18, 19, 25, 21, 16, 15, 19, 22, 24, 21, 18, 17, 15, 26 and 24.
We may use α= 0.05 level of significance.
Solution: Replacing each value greater than 20 with a plus (+) sign and each
value less than
187
Since calculated Z = -0.26 lies between Z = - 1.96 and Z = 1.96 (the critical
value of Z at 0.05 level of significance), the null hypothesis is accepted.
The two-sample sign test
Example 1: Suppose we have the following table indicating the ratings
assigned to two brands of cold drink X and Y by 12 consumers. Each respondent was
asked to taste the two brands of cold drink and then rate them.
Table 15.1. Ratings of brands X and Y cold drinks
Brand X 26 30 44 23 18 50 34 16 25 49 37 20
Brand Y 22 27 39 71 56 30 14 18 51 33 16
Sign + + + + + - + + + - + +
We have to apply the two-sample sign test. H0being both brands enjoy equal
preference.
Solution: Row three of Table 15.1 shows + and – signs. When X’s rating is
higher than that of Y, then the third row shows the ‘+’ sign. As against this, when
X’s rating is lower than that of Y, then it shows the ‘-‘ sign. The table shows 10 plus
signs and 2 minus signs. Now, we have to examine whether ’10 successes in 12
trials’ supports the null hypothesis p = ½ or the alternative hypothesis p > ½. The
null hypothesis implies that both the brands enjoy equal preferences and none is
better than the other. The alternative hypothesis is that the brand X is better than
brand Y. Referring to the binomial probabilities table, we find that for n = 12 and p
= ½ the probability of ’10 or more successes’ is 0.016 + 0.003 = 0.019. It follows
188
that the null hypothesis can be rejected at α= 0.05 level of significance. We can,
therefore, conclude that brand X is a preferred brand as compared to brand Y.
Example 2: To illustrate the second case, which relates to two independent samples, let
us consider the following data pertaining to the down times (periods in which computers
were inoperative on account of failures, in minutes of two different computers. We have
to apply the two-sample sign test.
Solution: These data are shown in Table 15.2 along with+ or – sign as may be
applicable in case of each pair of values. A plus sign is assigned when the downtime
for computer A is greater than that for computer B and a minus sign is given when
the downtime for computer
B is greater than that for computer A.
It will be seen that there are 15 plus signs and 7 minus signs. Thus, we have
to ascertain whether ‘15 successes in 22 trials’ support the null hypothesis p = ½.
The null hypothesis implies that the true average downtime is the same for both the
computers A and B. The alternative hypothesis is p ≠½. The null hypothesis implies
that the true average downtime is the same for both the computers A and B.The
alternative hypothesis is p = ½.
Let us use in this case the normal approximation of the binomial distribution.
This can be done since np and n (1 – p) are both equal to 11 in this example.
Substituting n = 22 and p = ½ into the formulas for the mean and the standard
deviation of the binomial distribution, we get µ = np = 22 (½) = 11 and σ= np (1 - p)
= 22.½.½ = 2.345
Hence, Z = (X – µ)/σ= (15 – 11)/2.345 = 1.71
Since this value of 1.71 falls between – Z0.025 = - 1.96 and Z0.025= 1.96, we
find that the null hypothesis cannot be rejected. This means thatthe downtime in
the two computers is the same.
189
Table 4.11.1 Friedman test for Mean Rank Variation of the Factors Which Motivates
Khadi Consumers to Buy Khadi Textile Products
420
N
Chi-Square 995.064
Df 7
The following data indicate the lifetime (in hours) of samples of two kinds of
light bulbs in continuous use:
Brand A 603 625 641 622 585 593 660 600 633 580 615 648
Brand B 620 640 646 620 652 639 590 646 631 669 610 619
We are required to use the Mann-Whitney test to compare the lifetimes of
brands A and B Light bulbs.
Solution: The first step for performing the Mann-Whitney test is to rank the
given data jointly (as if they were one sample) in an increasing or decreasing order
of magnitude. For our data, we thus obtain the following array where we use the
letters A and B to denote whether the light bulb was from brand A or brand B.
Table 1 Ranking of light bulbs of brands A and B
U = n1n2+ [n1(n1+ 1)]/2 – R1Where n1and n2are the sample sizes of Group 1
and Group 2, respectively, and R1is the sum of the ranks assigned to the values of
the first sample. In our example, we have n1= 12, n2= 12 and R1= 1 + 2 + 4 + 5 + 6
+ 8 + 12 + 13 + 15 + 18 + 21 + 23 = 128. Substituting these values in the above
formula,
U = (12) (12) + [12 (12 + 1)]/2 – 128
= 144 + 78 – 128
= 94
From Appendix Table 9 for n1and n2, each equal to 12, and for 0.05 level of
significance is 37. Since the critical value is smaller than the calculated value of 94,
we accept the null hypothesis and conclude that there is no difference in the
average lifetimes of the two brands of light bulbs. The test statistic we have just
applied is suitable when n1and n2are less than or equal to 25. For larger values of
n1and/or n2, we can make use of the fact that the distribution of Ws approaches a
normal distribution as sample sizes increase. We can then use the Z test to test the
hypothesis. The normal approximation
Although our observations are limited, we may apply the normal
approximation to this problem. For this, we have to use the Z statistic.
values. To use the Wilcoxon Mann Whitney test (a) It must be reasonable to regard
the data as a random sample from their respective populations (b) Observations
within each sample must be independent of one another (c) The two samples must
be independent of one another
Z -1.433
Asymp. Sig. (2-tailed) 0.152
The table above shows that majority of Khadi textile products are purchased
by male consumers (mean rank 207.21) than female consumers even if this mean
variation is statistically not significant at 5 per cent level od significance(P>.05 in
the above case).
Kruskal-Wallis Test
The Kruskal–Wallis one-way analysis of variance by ranks is a non-
parametric method for testing whether samples originate from the same
distribution. It is used for comparing two or more samples that are independent,
and that may have different sample sizes, and extends the Mann–Whitney Utest to
more than two groups. The parametric equivalent of the Kruskal -Wallis test is
the one-way analysis of variance (ANOVA). When rejecting the null hypothesis of
the Kruskal-Wallis test, then at least one of sample stochastically dominates at
least one other sample. The test does not identify where this stochastic dominance
occurs or for how many pairs of groups stochastic dominance obtains.It would help
to analyze the specific sample pairs for stochastic dominance. Since it is a non -
parametric method, the Kruskal–Wallis test does not assume a normal
distribution of the residuals, unlike the analogous one -way analysis of variance. If
the researcher can make the more stringent assumptions of an identically shaped
and scaled distribution for all groups, except for any difference in medians, then
the null hypothesis is that the medians of all groups are equal, and the alternative
hypothesis is that at least one population median of one group is different than the
194
population median of at least one other group. This test is used in this study to
analyze the preference of khadi consumers towards Khadi textile products based on
their place of residence (urban, semi -urban and rural) by assinging ranks to their
preferences.
This test is appropriate for use under the following circumstances: (a) when
there is three or more conditions that researcher want to compare; (b) each
condition is performed by a different group of parti cipants; i.e. researcher have an
independent-measures design with three or more conditions. (c) the data do not
meet the requirements for a parametric test. (i.e. use it if the data are not normally
distributed; if the variances for the different condition s are markedly different; or if
the data are measurements on an ordinal scale). If the data meet the requirements
for a parametric test, it is better to use one -way independent-measures Analysis of
Variance (ANOVA) because it is more powerful than the Kruskal-Wallis test.
F-test, used in analysis of variance, was that each of the k populations should
be normal with equal variance. In contrast to this, the Kruskal -Wallis test only
assumes that the k populations are continuous and have the same pattern
(symmetrical or skewed) of distribution. The null and the alternative hypotheses
ofthe Kruskal-Wallis test are:
H0: m1= m2= … = mk(i.e., means of the k populations are equal)
Ha: Not all mi’s are equal.
The Test Statistic: The computation of the test statisti cfollows a procedure that
is very similar to the Mann-Whitney Wilcoxon test.
(i) Rank all the n1+ n2+ … + nk= n observations, arrayed in ascending order.
(ii) Find R1, R2, … Rk, where Riis the sum of ranks of the ith sample. The test
statistic, denoted by H, is given by
Use the Kruskal-Wallis test to test the hypothesis that the effectiveness of the
three weight reducing diet plans is same at 5% level of significance.
195
Solution:
It is given that n1= 6, n2= 7 and n3= 9.
The total number of observations is 6 + 7 + 9 = 22. These are ranked in their
ascending order as given below:
The tabulated value of χ2at 2 d.f. and 5% level of significance is 5.99. Since H
is greater than this value, H0is rejected at 5% level of significance. e.g.). If the data
meet the requirements for a parametric test, it is better to use one -way
independent-measures Analysis of Variance (ANOVA) because it is more powerful
than the Kruskal-Wallis test.
LESSON – 15
CHI SQUARE TEST
15.1 INTRODUCTION
Any sample comprising numerical observations can be treated in the same
manner by using the letters a and b to denote, respectively, values above the
median and values below the median of the sample. In case an observation is equal
to the median, it is omitted. The resulting series of as and bs (representing the data
in their original order) can be tested for randomness on the basis of the total
number of runs above and below the median,respectively.
15.2 OBJECTIVE
After going through this Lesson you should be able to Understand Chi square
test, steps involved in chi square test and Goodness of fit.
15.3 CONTENT
15.3.1 Chi square test
15.3.2 Steps involved in chi square test
15.3.3 Goodness of fit
Let us take an example.
Example:
Suppose we have the following series of29 college students. After performing a
set of study exercises, increases in their pulse rate were recorded as follows:
22, 23, 21, 25, 33, 32, 25, 30, 17, 20, 26, 12, 21, 20, 27, 24, 28, 14, 29, 23,
22, 36, 25, 21, 23, 19, 17, 26 and 26. We have to test the randomness of these
data.
Solution:First, we have to calculate the median of this series. If we arrange
these values in an ascending order, we find that the size of (n +1)/2th item, that is,
15thitem is 24. Thus, the median is 24. As there is one value, which is 24 we omit it and
get the following arrangement of as and bs where a stands for an item greater than (or
above) the median and b stands for an item lower than (or below) the median:
bbb aaaaa bb a bbb aa b a bb aa bbbb aa
On the basis of this arrangement, we find that n1, (i.e. a) = 13, n2, (i.e. bs) =
15, and u = 12, we get
199
Since Z = -1.14 falls between –Z0.025= -1.96 and Z0.025= 1.96, the null
hypothesis cannot be rejected at the level of significance α= 0.05. We can, therefore,
conclude that the randomness of the original sample cannot be questioned. It may
be noted that this test is particularly useful in detecting trends and cyclic patterns
in a series. If there is a trend, there would be first mostly as and later mostly bs or
vice versa. In case of a repeated cyclic pattern, there would be a systematic
alternation of as and bs and probably, too many runs.
CHI-Square analysis
The Chi-square test is an important test amongst the several tests of
significance developed by statisticians. Chi -square, symbolically written as X 2, is a
statistical measure used in the context of sampling analysis for comparing a
variance to a theoretical variance. It is a non -parametric test, “it could be used to
determine if categorical data shows dependency or the two classifications are
independent. It can also be used to make comparisons between theoretical
populations and actual data when categories are used. Thus, the Chi -square test is
applicable in large number of problems. The test is a technique through the use of
which it is possible for all researchers to do the following 2.I) Test the goodness of fit,
ii) Test the significance of association between two attributes and i ii) Test the
homogeneity or the significance of population variance.
Steps involved in Applying Chi-Square Analysis
1. Calculate the expected frequencies on the basis of given hypothesis or on
the basis of null hypothesis.
2. Obtain the difference between observed and expected frequencies and find
out the squares of such differences i.e. calculate (Oij-Eij)2.
3. Divide the quantity (Oij-Eij)2 obtained as stated above by the corresponding
expected frequency to get (Oij-Eij)2 /Eij and this should be done for all the cell
frequencies or the group frequencies.
200
4. Find the summation of (Oij-Eij)2 / Eij values or what we call a (Oij-Eij)2 / Eij.
This is the required C2 value. The X2 value obtained as such should be compared
with relevant table value of X2 and then inference be drawn.
Goodness of Fit Test
This section begins with a short sketch of the method used to conduct achi
square goodness of fit test. An example of such a test is then provide d,followed by a
more detailed discussion of the methods used. Finall y, thereis a variety of examples
used to illustrate some of the common types ofgoodness of fit tests.The chi square
goodness of fit test begins by hypothesizing that the distribution of a variable
behaves in a particular manner. For example, inorder to determine daily staffing
needs of a retail store, the manager maywish to know whether there are an equal
number of customers each day ofthe week. To begin, an hypothesis of equal
numbers of customers on eachday could be assumed, and this would be the null
hypothesis. A student may observe the set of grades for a class, and suspect that
the professor allocatedthe grades on the basis of a normal distribution. Another
possibility is thata researcher claims that the sampleselected has a distribution
which is veryclose to distribution of the population. While no general statement can
beprovided to cover all these possibilities, what is common to all is that aclaim has
been made concerning the nature of the whole distribution.Suppose that a variable
has a frequency distribution with k categoriesinto which the data has been
grouped. The frequencies of occurrence of thevariable, for each category of the
variable, are called the observed values.
The manner in which the chi square goodness of fit test works is to
determinehow many cases there would be in each category if the sample data
weredistributed exactly according to the claim. These are termed the
expectednumber of cases for each category. The total of the expected number
ofcases is always made equal to the total of the observed number of cases.
The null hypothesis is that the observed number of cases in each categoryis
exactly equal to the expected number of cases in each category. Thealternative
hypothesis is that the observed and expected number of casesdiffer sufficiently to
reject the null hypothesis.
Let Oiis the observed number of cases in category i and Eiis the expected
number of cases in each category, for each of the k categoriesi = 1, 2, 3, · · · , k ,
into which the data has been grouped. The hypotheses are
H0: Oi = Ei
H1: Oi # Ei and the test statistic is
where the summation proceeds across all k categories, and there are k −
1degrees of freedom.Large values of this statistic lead the researcher to rejectthe
201
null hypothesis, smaller values mean that the null hypothesis cannotbe rejected.
The chi square distribution with k − 1 degrees of freedom inAppendix ?? is used to
determine the critical χ2value for the test.This sketch of the test should allow you
to follow the first example of achi square goodness of fit test. After Example 10.3.1,
the complete notationfor a chi square goodness of fit test is provided. This notation
will becomemore meaningful after you have seen how a goodness of fit test is
conducted.
Median test for two independent samples
In order to perform this test, let us use our previous example, which pertains
to the Down times of the two computers. The median of the combined data is 52,
which can easily be checked. There are 5 values below 52 and 15 values above it, in
case of computer A. As regards computer B, the corresponding figures are 16 and
6. All this information is Summarized in Table 15.3, which also indicates the totals
of the rows and columns.
Example 1
Table 1. Classification of downtime for computers A and B
Our null hypothesis H0is that there is no difference inthe median downtime for
the two computers. The alternative hypothesis H1is that there is difference in the
downtime of the two computers.
We now calculate the expected frequencies by the formula (Rowi ×
Columni)/Grand total. Thus, Table 4 shows both the observed and the expected
frequencies. Of course, we could have obtained these results by arguing that half
the values in each sample can be expected to fall above the median and the other
half below it.
Table 2. Calculation of chi-square
202
TABLE - 4.2.2(A)
CHI-SQUARE TEST
DF S/NS Remarks
Pearson Chi-Square Calculated Chi-square Value Table value
Test whether the differences among the three sample means are significant.
Solution: In case of such problems, analysis of variance is ordinarily
performed. However, here we find that the data for Section C have much more
variability as compared to the data for the other two sections. In view of this, it
would be wrong to assume that the three population standard deviations are the
same. This means that the method of one-way analysis of variance cannot be used.
In order to perform a median test, we should first determine the median of the
combined data. This comes out to 63.5, as can easily be checked. Then we count
how many of the marks in each sample fall below or above the median. Thus, the
results obtained are shown in Table 1
204
Since the corresponding expected frequencies for Section A are 10 and 10, for
Section B are 8 and 8, and for Section C 10 and 10, we can obtain the value of chi-
square. These calculations are shown below:
Now, we have to compare this value with the critical value of χ2at 5 per cent
level of significance. This value is 5.991 for 2 (K – 1 = 3 – 1) degrees of freedom (Chi-
square Table). As the calculated value of χ2is less than the critical value, the null
hypothesis cannot be rejected. In other words, we cannot conclude that there is a
difference in the true average (median) marks obtained by the students in Statistics
test from the three sections.
EXAMPLE
SIZE OF FARMER AND LEVEL OF SATISFACTION TOWARDS PRODUCTION OF
BANANA: (TWO WAY TABLE)
Level of satisfaction Total
HDS DS N S HS
No.of.Respondents
Small 42 42 28 16 22 150
Percentage 28.0% 28.0% 18.7% 10.7% 14.7% 100.0%
SIZE OF No.of.Respondents 12 19 20 30 19 100
FARMERS medium Percentage 12.0% 19.0% 20.0% 30.0% 19.0% 100.0%
No.of.Respondents 6 7 10 14 13 50
Large
Percentage 12.0% 14.0% 20.0% 28.0% 26.0% 100.0%
No.of.Respondents 60 68 58 60 54 300
Total
Percentage 20.0% 22.7% 19.3% 20.0% 18.0% 100.0%
Source: Primary Data
Null hypothesis: There is no relationship between the size of farmers and
level of satisfaction towards production of banana among the small, medium, large
farmers.
Alternative hypothesis: There is a relationship between the size of farmers
level of satisfaction towards production of banana among the small, medium, large
farmers.
The relationship between size of farmers and level of satisfaction towards the
production of banana. Out of the 300 respondents, 150 respondent up to three
members in these factions, 14.7 percent of the respondents fall under highly
satisfied and 28 percent of the respondents fall under highly dissatisfied about the
205
Brand
6 2327 3.10 3119 4.16 3110 4.15 3096 4.13 3171 4.23
Name
Name of
the
alliance
7 1792 2.39 1762 2.35 2456 3.27 2711 3.61 2330 3.11
or non-
profit
partner
Source: Primary Data
15.4 REVISION POINTS
1. Chi square test
2. Steps involved in chi square test
3. Goodness of fit
15.5 INTEXT QUESTION
Does the new math program improve student performance? Suppose you take
a random sample of 20 students who are using a new algebra text which features
group work and unit summaries and a second sample of 30 students who are using
a more traditional text. You compare student achievement on the state test given to
all students at the end of the course. Use the frequency table to determine if the
proportions from each group are equal at each performance level.
Support: No Support:
Conform: 18 40
Not Conform: 32 10
Is there a significant difference between the "support" and "no support"
conditions in the frequency with which individuals are likely to conform?
15.10 SUGGESTED READINGS / REFERENCE BOOKS
1. Ackoff, Russell L., The Design of Social Research, Chicago: University of
Chicago Press, 1961.
2. Ackoff, Russell L., Scientific Method, New York: John Wiley & Sons, 1962.
3. Allen, T. Harrell, New Methods in Social Science Research, New York:
Praeger Publishers, 1978.
4. Anderson, H.H., and Anderson, G.L., An Introduction to Projective
Techniques and Other Devices for Understanding the Dynamics of Human
Behaviour, New York: Prentice Hall, 1951.
5. Anderson, T.W., An Introduction to Multivariate Analysis, New York: John
Wiley & Sons, 1958.
6. Bailey, Kenneth D., “Methods of Social Research,” New York, 1978.
7. Baker, R.P., and Howell, A.C., The Preparation of Reports, New York: Ronald
Press, 1938.
8. Bartee, T.C., “Digital Computer Fundamentals,” 5th Ed., McGraw-Hill,
International Book Co., 1981.
9. Barzun, Jacques, and Graff, Henery, F., The Modern Researcher, rev. ed.,
New York: Harcourt, Brace & World, Inc., 1970.
10. Bell, J.E., Projective Techniques: A. Dynamic Approach to the Study of
Personality, New York: Longmans Green, 1948.
11. Bellenger, Danny N., and Greenberg, Barnett A., Marketing Research—A
Management Information Approach, Homewood, Illinois: Richard D. Irwin, Inc.,
1978.
15.11 LEARNING ACTIVITIES
A researcher is interested in sex differences in social play in cats. She observes
sixteen kittens (eight male and eight females), and records the sex of the
participants in 500 play-bouts, taking a note of the sex of the animal initiating a
bout and the sex of the animal who responds to the initiation. Are there sex -
differences in the kittens' play? Perform a Chi -Square test on these data. Is there
anything wrong with doing this test?
Respondent Male Female
initiator of bout: Male: 200 150
Female: 100 50
15.12 KEY WORDS
Chi square test, steps involved in Chi square test and Goodness of fit.
208
LESSON – 16
MANN-WHITNEY TEST
16.1 INTRODUCTION
Statistics is a set of tools used to organize and analyze data. Data must either
be numeric in origin or transformed by researchers into numbers. Employing
statistics serves two purposes, (1) description and (2) prediction. Statistics are used
to describe the characteristics of groups. These characteristics are referred to as
variables. Data is gathered and recorded for each variable. Descriptive statistics can
then be used to reveal the distribution of the data in each variable. Statistics is also
frequently used for purposes of prediction. Prediction is based on the concept of
generalizability: if enough data is compiled about a particular context (e.g.,
students studying writing in a specific set of classrooms), the patterns revealed
through analysis of the data collected about that context can be generalized (or
predicted to occur in) similar contexts. The prediction of what will happen in a
similar context is probabilistic. That is, the researcher is not certain that the same
things will happen in other contexts; instead, the researcher can only reasonably
expect that the same things will happen.
16.2 OBJECTIVE
The main aim of this Lesson is to study the Basics of statistics. After going
through this Lesson you should be able to:
Describe the origin of Statistics
Explain about the importance of Statistics
Use this test when two different groups of participants perform both conditions
of the study: i.e., it is ppropriate for analysing the data from an independent-
measures design with two conditions. Use it when the data do not meet the
requirements for a parametric test (i.e. if the data are not normally distributed; if
the variances for the two conditions are markedly different; or if the data are
measurements on an ordinal scale). Otherwise, if the data meet the requirements
for a parametric test, it is better to use an independent-measures t-test (also known
as a "two-sample" t-test). The logic behind the Mann-Whitney test is to rank the
data for each condition, and then see how different the two rank totals are. If there
is a systematic difference between the two conditions, then most of the high ranks
will belong to one condition and most of the low ranks will belong to the other one.
As a result, the rank totals will be quite different. On the other hand, if the two
conditions are similar, then high and low ranks will be distributed fairly evenly
between the two conditions and the rank totals will be fairly similar. The Mann -
Whitney test statistic "U" reflects the difference between the two rank totals. The
SMALLER it is (taking into account how many participants you have in each group)
then the less likely it is to have occurred by chance. In this study this test is used
to know the rank preference of Khadi consumers based on gender wise
(male/female) towards Khadi textile products.
209
16.3 CONTENT
16.3.1 Meaning
16.3.2 Definition
16.3.3 Uses of statistics
16.3.1 MEANING
“Statistics”, that a word is often used, has been derived from the Latin word
‘Status’ that means a group of numbers or figures; those represent some
information of our human interest. We find statistics in everyday life, such as in
books or other information papers or TV or newspapers. Although, in the beginning
it was used by Kings only for collecting information about states and other
information which was needed about their people, their numbe r, revenue of the
state etc. This was known as the science of the state because it was used only by
the Kings. So it got its development as ‘Kings’ subject or ‘Science of Kings’ or we
may call it as “Political arithmetic’s”. It was for the first time, perhaps in Egypt to
conduct census of population in 3050 B.C. because the king needed money to erect
pyramids. But in India, it is thought, that, it started dating back to Chandra Gupta
Maurya’s kingdom under Chankya to collect the data of births and deaths. T M has
also been stated in Chankya’s Arthshastra. But now-a-days due to its pervading
nature, its scope has increased and widened. It is now used in almost in all the
fields of human knowledge and skills like Business, Commerce, Economics, Social
Sciences, Politics, Planning, Medicine and other sciences, Physical as well as
Natural.
16.3.2 DEFINITION
The term ‘Statistics’ has been defined in two senses, i.e. in Singular and in
Plural sense.
a) In the Plural Sense
“Statistics are numerical statements of facts in any department of enquiry
placed in relation to each other.” —A.L. Bowley
“The classified facts respecting the condition of the people in a state —
especially those facts which can be stated in numbers or in tables of numbers or in
any tabular or classified arrangement.” —Webster
“By statistics we mean aggregates of facts affected to a marked extent by
multiplicity of causes, numerically expressed, enumerated or estimated according
to reasonable standard of accuracy, collected in a systematic manner for a
predetermined purpose, and placed in relation to each other.” —Horace Sacrist
b) In the Singular Sense
“Statistics refers to the body of technique or methodology, which has been
developed for the collection, presentation and analysis of quantitative data and for
the use of such data in decision making.” —Ncttor and Washerman
“Statistics may rightly be called the science of averages.” —Bowleg
210
the concept of Population. It has be en argued that since statistics usually involves
the study of human behavior, it cannot claim the precision of the physical sciences.
Probability has much longer history. Probability is derived from the verb to
probe meaning to "find out" what is not too easily accessible or understandable .
The word "proof" has the same origin that provides necessary details to understand
what is claimed to be true.
Probability originated from the study of games of chance and gambling during
the 16 th century. Probability theory was a branch of mathematics studied by Blaise
Pascal and Pierre de Fermat in the seventeenth century. Currently in 21 st century,
probabilistic modeling is used to control the flow of traffic through a highway
system, a telephone interchange, or a computer processor; find the genetic makeup
of individuals or populations; quality control; insurance; investment; and other
sectors of business and industry.
New and ever growing diverse fields of human activities are using statistics;
however, it seems that this field itself remains obscure to the public. Professor
Bradley Efron expressed this fact nicely:
During the 20 th Century statistical thinking and methodology have become the
scientific framework for literally dozens of fields including education, agriculture,
economics, biology, and medicine, and with increasing influence recently on the
hard sciences such as astronomy, geology, and physics. In other words, we have
grown from a small obscure field into a big obscure field.
GARRETT RANKING TECHNIQUE
This technique was used to rank the preference of the respondents on different
aspects of the study. The order of merit given by the respondents were converted
into ranks by using the following formula.
100 (R 0.5)
ij
Percentage Position =
H
j
The percentage position of each rank thus obtained is converted into sc ores by
referring to the table given by Henry Garrett. Then for each factor the scores of
individual respondents are added together and divided by the total number of
respondents for whom the scores were added. These mean scores for all the factors
were arranged in the descending order, ranks given and most important aspects
identified.
212
position of the data in a rank ordering, unlike the t test which uses the actual
values. To use the Wilcoxon Mann Whitney test (a) It must be reasonable to regard
the data as a random sample from their respective populations (b) Observations
within each sample must be independent of one another (c) The two samples must
be independent of one another
The following data indicate the lifetime (in hours) of samples of two kinds of
light bulbs in continuous use:
Brand A 603 625 641 622 585 593 660 600 633 580 615 648
Brand B 620 640 646 620 652 639 590 646 631 669 610 619
We are required to use the Mann-Whitney test to compare the lifetimes of
brands A and B
Light bulbs
Solution:The first step for performing the Mann-Whitney test is to rank the
given data jointly(as if they were one sample) in an increasing or decreasing order of
magnitude. For our data, we thus obtain the following array where we use the
letters A and B to denote whether the light bulb was from brand A or brand B.
Table 1 Ranking of light bulbs of brands A and B
Sample score Group Rank Sample score Group Rank
580 A 1 625re A 13
585 A 2 631 B 14
590 B 3 633 A 15
593 A 4 639 B 16
600 A 5 640 B 17
603 A 6 641 A 18
610 B 7 646 B 19.5
615 A 8 646 B 19.5
619 B 9 648 A 21
620 B 10.5 652 B 22
620 B 10.5 660 A 23
622 A 12 669 B 24
U = n1n2+ [n1(n1+ 1)]/2 – R1Where n1and n2are the sample sizes of Group 1
and Group 2, respectively, and R1is the sum of the ranks assigned to the values of
the first sample. In our example, we have n1= 12, n2= 12 and R1= 1 + 2 + 4 + 5 + 6
+ 8 + 12 + 13 + 15 + 18 + 21 + 23 = 128. Substituting these values in the above
formula,
U = (12) (12) + [12 (12 + 1)]/2 – 128
= 144 + 78 – 128
= 94
From Appendix Table 9 for n1and n2, each equal to 12, and for 0.05 level of
significance is 37. Since the critical value is smaller than the calcul ated value of 94,
we accept the null hypothesis and conclude that there is no difference in the
average lifetimes of the two brands of light bulbs. The test statistic we have just
applied is suitable when n1and n2are less than or equal to 25. For larger values of
n1and/or n2, we can make use of the fact that the distribution of Ws approaches a
normal distribution as sample sizes increase. We can then use the Z test to test the
hypothesis. The normal approximation
Although our observations are limited, we may apply the normal approximation
to this problem. For this, we have to use the Z statistic.
n1 n2 (n1 n2 1
2. Standard error =
12
12 12 (12 12) 1
300 17.3
12
3. (Statistic – Mean)/Standard deviation
= (94 – 72)/17.3 = 1.27
The critical value of Z at 0.05 level of significance is 1.64. Since the calculated
value of Z = 1.27 is smaller than 1.64, the null hypothesis is accepted. This shows
that there is no difference in average lifeti mes of brands A and B bulbs. The Z test
is more dependable as compared to the earlier one. It may be noted that Mann -
Whitney test required fewer assumptions than the corresponding standard test. In
fact, the only assumption required is that the populations from which samples have
been drawn are continuous. In actual practice, even when this assumption turns
out to be wrong, this is not regarded serious.
e.g:2
In the independent samples case, it is preferable to apply the independent
sample t test when there exists normality of the sample mean for each sample. If
there is a violation of this assumption, then a non parametric test for independent
samples is appied – the Wilcoxon Mann Whitney test. This test uses the relative
216
position of the data in a rank ordering, unlike the t test which uses the actual
values. To use the Wilcoxon Mann Whitney test (a) It must be reasonable to regard
the data as a random sample from their respective populations (b) Observations
within each sample must be independent of one another (c) The two samples must
be independent of one another
Gender wise Mean Ranks of Khadi Textile Products
Mean Sum of
Gender N Rank Ranks
218
UNIT – V
LESSON – 17
variable (also known as the outcome or phenomenon of interest), more than one
independent variable (also known as a predictor) or both. Upper-level
undergraduate courses and graduate courses in statistics teach multivariate
statistical analysis. This type of analysis is desirable because researchers often
hypothesize that a given outcome of interest is effected or influenced by more than
one thing.
16.2 OBJECTIVE
After going through this Lesson you should be able to understand the types of
multivariate analysis, multiple regression, factor analysis, Path analysis and
MANOVA
17.3 CONTENT
17.3.1 Types
17.3.2 Regression Analysis
17.3.3 Factor analysis
17.3.4 Path analysis
17.3.5 MANOVA
17.3.1 Types
There are many statistical techniques for conducting multivariate analysis,
and the most appropriate technique for a given study varies with the type of study
and the key research questions. Four of the most common multivariate techniques
are multiple regression analysis, factor analysis, path analysis and multiple
analysis of variance, or MANOVA.
17.3.2 Multiple Regressions
Multiple regression analysis, often referred to simply as regression analysis,
examines the effects of multiple independent variables (predictors) on the value of a
dependent variable, or outcome. Regression calculates a coefficient for each
independent variable, as well as its statistical significance, to estimate the effect of
each predictor on the dependent variable, with other predictors held constant.
Researchers in economics and other social sciences often use regression analysis to
study social and economic phenomena. An example of a regression study is to
examine the effect of education, experience, gender, and ethnicity on income.
17.3.3 Factor Analysis
Factor analysis is a data reduction technique in which a researcher reduces a
large number of variables to a smaller, more manageable, number of factors. Factor
analysis uncovers patterns among variables and then clusters highly interrelated
variables into factors. Factor analysis has many applications, but a common use is
in survey research, where researchers use the technique to see if lengthy series of
questions can be grouped into shorter sets.
17.3.4 Path Analysis
This is a graphical form of multivariate statistical analysis in which graphs
known as path diagrams depict the correlations among variables, as well as the
220
directions of those correlations and the "paths" alon g which these relationships
travel. Statistical software programs calculate path coefficients, the values of which
estimate the strength of relationships among the variables in a researcher's
hypothesized model.
The term ‘path analysis’ was first introduce d by the biologist Sewall Wright in
1934 in connection with decomposing the total correlation between any two
variables in a causal system. The techniqueof path analysis is based on a series of
multiple regression analyses with the added assumption ofcausal relationship
between independent and dependent variables. This technique lays relativelyheavier
emphasis on the heuristic use of visual diagram, technically described as a path
diagram. Anillustrative path diagram showing interrelationships between Fathe rs’
education, Fathers’ occupation,Sons’ education, Sons’ first and Sons’ present
occupation can be shown in the Fig. 13.2.
Path analysis makes use of standardized partial regression coefficients (known
as beta weights) as effect coefficients. In linear additive effects are assumed, then
through path analysis a simple setof equations can be built up showing how each
variable depends on preceding variables. “The mainprinciple of path analysis is that
any correlation coefficient between two variables, or a gross oroverall measure of
empirical relationship can be decomposed into a series of parts: separate paths
ofinfluence leading through chronologically intermediate variable to which both the
correlated variables have links.”
The merit of path analysis in comparison to correlational analysis is that it
makes possible theassessment of the relative influence of each antecedent or
explanatory variable on the consequent orcriterion variables by first making explicit
the assumptions underlying the causal connection s andthen by elucidating the
indirect effect of the explanatory variables.
“The use of the path analysis technique requires the assumption that there
are linear additive, a symmetric relationships among a set of variables which can be
measured at least on a quasi-intervalscale. Each dependent variable is regarded as
determined by the variables preceding it in the pathdiagram, and a residual
221
x1 e
x p X1 e pX e
2 21 1 2
x 3 p31X2 X2 e3
where the Xvariables are measured as deviations from their respective means.
p21may be estimated from the simple regression of X 2on X1i.e., X2 = b21 Xl and p31
and p32may be estimated from the regression of X 3 on X2 and X1 as under:
X3 b31.2 X1 b21 X2
where b31.2 means the standardized partial regression coefficient for predicting
variable 3 from variable 1 when the effect of variable 2 is held constant.
In path analysis the beta coefficient i ndicates the direct effect of Xj (j = 1, 2, 3,
..., p) on the dependent variable. Squaring the direct effect yields the proportion of
the variance in the dependentvariable Ywhich is due to each of the pnumber of
independent variables Xj (i = 1, 2, 3, ..., p). After calculating the direct effect, one
may then obtain a summary measure of the total indirect effect ofXjon the
dependent variable Yby subtracting from the zero correlation coefficient ryxj, the
beta coefficient bji.e.,
Such indirect effects include the unanalysed effects and spurious relationships
due to antecedent variables. In the end, it may again be emphasised that the main
virtue of path analysis lies in making explicitthe assumptions underlying the causal
connections and in elucidating the indirect effects due toantecedent variables of the
given system.
17.3.5 MANOVA
Multiple Analysis of Variance, or MANOVA, is an advanced form of the more
basic analysis of variance, or ANOVA. MANOVA extends the technique to studies
with two or more related dependent variables while controlling for the correlations
among them. An example of a study for which MANOVA would be an appropriate
technique is a study of health among three groups of teens: those who exercise
222
regularly, those who exercise on occasion, and those who never exercise. A
MANOVA for this study would allow multiple health-related outcome measures
such as weight, heart rate, and respiratory rates.
Benefits
Multivariate statistical analysis is especially important in social science
research because researchers in these fields are often unable to use randomized
laboratory experiments that their counterparts in medicine and natural sciences
often use. Instead, many social scientists must rely on quasi -experimental designs
in which the experimental and control groups may have initial differences that
could affect or bias the outcome of the study. Multivariate techniques try to
statistically account for these differences and adjust outcome measures to control
for the portion that can be attributed to the differences.
Example
A research team wants to test the user acceptance with a new onli ne travel
booking tool. The team conducts a study where they assign 30 randomly chosen
people into 3 groups. The first group needs to book their travel through an
automated online-portal; the second group books over the phone via a hotline; the
third group sends a request via the online-portal and receives a call back. The
team measures user acceptance as the behavioral intention to use the system, they
do measure the latent construct behavioral intention with 3 variables – ease of use,
perceived usefulness, effort to use.
In the example, some statisticians argue that the MANOVA can only find the
differences in the behavioral intention to use the system. However, some
statisticians argue that you can establish a causal relationship between the
channel they used and the behavioral intention for future use. It is generally
assumed that the MANOVA is an ‘analysis of dependencies’. It is referred to as
such because it proves an assumed cause -effect relationship between two or more
independent variables and two or more dependent variables. In more statistical
terms, it tests the effect of one or more independent variables on one or more
dependent variables.
Other things you may want to try
When faced with a question similar to the one in our example you cou ld also
try to run a 3 factorial ANOVAs, testing the influence of the three independent
variables (the three channels) on each of the three dependent variables (ease of use,
perceived usefulness, effort to use) individually. However running multiple factorial
ANOVAs does not account for the full variability in all three dependent variables
and thus the test has lesser power than the MANOVA.
Another thing you might want to try is running a factor analysis on the three
dependent variables and then running a factorial ANOVA. The factor analysis
reduces the variance within the three dependent variables to one factor, thus this
procedure does have lesser power than the MANOVA.A third approach would be to
conduct a discriminant analysis and switch the dependent and independent
223
variables. That is the discriminant analysis uses the three groups (online, phone,
call back) as the dependent variable and identifies the significantly discriminating
variables from the list of continuous-level variables (ease of use, perceived
usefulness, effort to use).
Mathematically, the MANOVA is fully equivalent to the discriminant analysis.
The difference consists of a switching of the independent and dependent variables.
Both the MANOVA and the discriminant analysis are a series of canonical
regressions. The MANOVA is therefore the best test use when conducting
experiments with latent variables. This is due to the fact that it only requires a
nominal scale for the independent variables which typically represent the
treatment. This includes multiple continuous-level independent variables – which
typically measure one latent (not directly observable) construct.
The MANOVA is much like the one-way ANOVA and the factorial ANOVA in
that the one-way ANOVA has exactly one independent and one dependent variable.
The factorial ANOVAs can have one or more independent variables but always has
only one dependent variable. On the other hand the MANOVA can have two or
more dependent variables.
Considerations
Statistical software programs such as SAS, Stata, and SPSS can perform
multivariate statistical analyses. These programs are frequently used by university
researchers and other research professionals. Spreadsheet programs can perform
some multivariate analyses, but are intended for more g eneral use and may have
limited abilities than a specialized statistical software package.
One-Way MANOVA
MANOVA is short for Multivariate ANalysis Of Variance. The main purpose of
a one-way ANOVA is to test if two or more groups differ from each other
significantly in one or more characteristics. A factorial ANOVA compares means
across two or more variables. Again, a one-way ANOVA has one independent
variable that splits the sample into two or more groups whereas the factorial
ANOVA has two or more independent variables that split the sample in four or more
groups. A MANOVA now has two or more independent variables and two or more
dependent variables.
For some statisticians the MANOVA doesn't only compare differences in mean
scores between multiple groups but also assumes a cause effect relationship
whereby one or more independent, controlled variables (the factors) cause the
significant difference of one or more characteristics. The factors sort the data
points into one of the groups causing the difference in the mean value of the
groups.
FACTOR ANALYSIS
Factor Analysis is a set of technique which by analyzing correlations between
variables reduces their numbers into fewer factors which explain much of the
original data, more economically. Even though a subjective interpretation can result
224
from a factor analysis output, the procedure often provides an insight into relevant
psychographic variables, and results in economic use of data collection efforts. The
subjective element of factor analysis is reduce d by splitting the sample randomly
into two and extracting factors separately from both parts. If similar factors result,
the analysis is assumed as reliable or stable.
A. METHODS OF CONDUCTING FACTOR ANALYSIS5, 6, 7
factor analysis is reduced by splitting the sample randomly into two and
extracting factors separately from both parts. If similar factors result, the
analysis is assumed as reliable or stable.
TABLE - 4.4.1.1
KMO AND BARTLETT’S TEST FOR FACTORS RELATED TO INFLUENCE IN
CHOOSING CRM PRODUCTS
Kaiser-Meyer-Olkin Measure of Sampling Adequacy .681
Bartlett’s Test of Sphericity: Approx. Chi -Square 2130.534
Sig 0.000**
S/NS S
**p<0.01 S-Significant
From the above table, two tests, namely Kaiser-Meyer-Olkin Measure of
Sampling Adequacy (KMO) and Bartlett’s Test of Sphericity have been applied to
test whether the relationship among the variables has been significant or not. The
Kaiser-Meyer-Olkin Measure of sampling adequacy shows that the value of test
statistics is 0.681, which means the factor analysis for the selected variable is
found to be appropriate or good to the data. Bartlett’s test of sphericity is used to
test whether the data are statistically significant or not with the value of test
statistics and the associated significance level. It shows that there exists a high
relationship among variables.
Table – 4.4.1.2
EIGEN VALUES AND PROPORTION OF TOTAL VARIANCE OF EACH NDERLYING
FACTORS RELATED TO INFLUENCE IN CHOOSING CRM PRODUCTS
Initial Eigen v alues Ex traction Sums of Squared Loadings Rotation Sums of Squared loadings
Component Total % of Cumulativ e % Total % of Cumulativ e Total % of Cumulativ e %
Variance Variance % Variance
1 3.418 28.486 28.486 3.418 28.486 28.486 2.080 17.331 17.331
2 1.570 13.081 41.567 1.570 13.081 41.567 1.881 15.679 33.010
3 1.254 10.450 52.018 1.254 10.450 52.018 1.854 15.449 48.459
4 1.070 8.915 60.933 1.070 8.915 60.933 1.497 12.474 60.933
The above table reveals the factor correlation matrix. If the factors are
uncorrelated among themselves, then in the factor correlation matrix, the diagonal
elements will be 1’s and off diagonal elements will be 0’s. Since matrix was rotated
with Varimax, barring some variables all other variables are found to have, even i f
not zero correlations but fairly, low correlation.
CHART
Component Plot228
in Rotated Space
1.0
Q3.1.1
Q3.1.3
Component 2
0.5 Q3.1.2
Q3.1.4 Q3.1.5
Q3.1.8
Q3.1.10
0.0 Q3.1.6
Q3.1.7 Q3.1.9
Q3.1.11
-0.5 Q3.1.12
-1.0
-1.0
-1.0 -0.5 -0.5
0.0
0.0 0.5
0.5 1.0 1.0 t3
Componen p onen
t1 Com
CONCLUSION
Thus the twelve variables in the data were reduced to four Component factors
and each factor may be identified with the corresponding variables as follows:
TABLE -4.4.1.6
SHOWING THE FACTORS IDENTIFIED THE FACTORS RELATED TO
INFLUENCE IN FOR CHOOSING CRM PRODUCTS
X6 CRM Campaign 50.84 %
X5 Brand familiarity 41.99%
Factor I
X12 Cause- Corporate- Brand fit 41.86%
X9 Contribution level to cause 40.83%
X1 Price of the product 54.91%
X2 Quality of the product 41.34%
Factor II
X3 Nature of Cause 37.58%
X4 Availability of the product 28.30%
X11 Sales promotional activities 62.88%
Factor III
X10 Attractive sales offers 49.42%
X7 Previous experience 70.56%
Factor IV
X8 Package of the product 34.69%
17.4 REVISION POINTS
229
1. Types
2. Multiple Regression Analysis
3. Factor analysis
4. Path analysis
5. MANOVA
17.5 INTEXT QUESTIONS
1. List out the advantages and disadvantages of Factor analysis
2. Explain the limitations of Path Analysis
17.6 SUMMARY
After going through this Lesson you should be able to explain the types of
multivariate analysis, multiple regression, factor analysis, Path analysis and
MANOVA
17.7 TERMINAL EXERSICE
1. Explain the importance of factor analysis in research
2. Explain the importance of Multiple Regression analysis in research
17.8 SUPPLEMENTARY MATERIALS
1. Barzun, Jacques, and Graff, Henery, F., The Modern Researcher, rev. ed.,
New York: Harcourt, Brace & World, Inc., 1970.
17.9 ASSIGNMENT
1. Briefly Explai how to conduct and interpret Factor analysis?
17.10 SUUGESTED READINGS/REFERENCE BOOKS
1. Ackoff, Russell L., The Design of Social Research, Chicago: University of
Chicago Press, 1961.
2. Ackoff, Russell L., Scientific Method, New York: John Wiley & Sons, 1962.
3. Allen, T. Harrell, New Methods in Social Science Research, New York:
Praeger Publishers, 1978.
4. Anderson, H.H., and Anderson, G.L., An Introduction to Projective
Techniques and Other Devices for Understanding the Dynamics of Human
Behaviour, New York: Prentice Hall, 1951.
5. Anderson, T.W., An Introduction to Multivariate Analysis, New York: John
Wiley & Sons, 1958.
6. Bailey, Kenneth D., “Methods of Social Research,” New York, 1978.
7. Baker, R.P., and Howell, A.C., The Preparation of Re ports, New York: Ronald
Press, 1938.
8. Bartee, T.C., “Digital Computer Fundamentals,” 5th Ed., McGraw-Hill,
International Book Co., 1981.
9. Barzun, Jacques, and Graff, Henery, F., The Modern Researcher, rev. ed.,
New York: Harcourt, Brace & World, Inc., 1970.
230
231
LESSON – 18
KRUSKAL-WALLIS TEST
18.1 INTRODUCTION
Respondent’s opinion towards factors influencing for choosing Cause related
marketing products. In the study area out of seven hundred and fifty respondents
were divided into two groups
18.2 OBJECTIVE
After going through this Lesson you should be able to explain the types of
Discriminant function analysis Canonical Discriminant function, Relative
Discriminant function
18.3 CONTENT
1. Discriminant function analysis
2. Canonical Discriminant function,
3. Relative Discriminant function
DISCRIMINANT FUNCTION ANALYSIS
Low level of factors influencing for choosing Cause related marketing products
and the high level of factors influencing for choosing Cause related marketing
products. The difference of opinion of the respondents in one group from the other
is studied with the help of discriminant function analysis. For the purpose of the
study, the following variables were selected.
1. Gender
2. Age
3. Marital status
4. Educational Qualification
5. Occupational status
6. Monthly Income
7. Family size
8. Nature of the family
9. Number of earning members in the family
10. Name of the Zone
The discriminant function analysis attempts to construct a function with these
and other variables so that the respondents belonging to these two groups are
differentiated at the maximum. The linear combination of variables is known as
discriminant function and its parameters are called discriminant function
coefficients. In constructing this discriminant function all the variables which
contribute to differentiate these three groups are examined.
Mahalanobis minimum D2 method is based on the generalized squared
Euclidean distance that adjusts for unequal variances in the variables. The major
232
The overall stepwise D.F.A results after all significant discriminators have been
included in the estimation of discriminated function are given in the following table .
TABLE -4.4.3.2
SUMMARY TABLE BETWEEN LOW LEVEL AND HIGH LEVEL GROUPS
Step Variables entered Wilk’s Lamda F-value Significance
1 Gender 0.897 85.729 .000**
2 Marital status 0.988 8.989 .003**
3 Educational Qualification 0.970 22.898 .000**
4 Family size 0.978 17.014 .000**
5 Nature of the family 0.976 18.740 .000**
*Significant at 1% level
The summary table indicates that variable gender entered in step one. The
variables such as Gender, Marital status, Educ ational Qualification, Family size
and Nature of the family are significant at one per cent significance level. All the
variables are significant discriminators based on their Wilk’s Lambda and F-value.
The multivariate aspect of this model is given in the following table.
TABLE -4.4.3.3
CANONICAL DISCRIMINANT FUNCTION
(BETWEEN LOW AND HIGH GROUPS)
Canonical S/NS
Wilk's Lamda Chi -square p-value
correlation
0.420 .823 144.438 .000** S
-0.459 0 +0.466
Educational
3.46 3.68 .506 0.11 13.16
Level
Family size 2.22 1.93 -.195 0.006 0.72
Nature of the
1.55 1.70 1.011 0.15 17.94
family
TOTAL 100
2 R 2 2
12 R1 2
R
k
H ... - 3 (n 1)
n(n 1) n n n
1 2 k
It can be shown that the distribution of H is c2with k – 1 d.f., when size of
each sample is at least 5. Thus, if H > 21 k− χ , H0 is rejected.
Example 1: To compare the effectiveness of three types of weight-reducing
diets, a homogeneous groups of 22 women was divided into three sub-groups and
each sub-group followed one of these diet plans for a period of two months. The
weight reductions, in kgs, were noted as given below:
I 12.5 9 6.5 17 14 11 70
Diet II 15 20 21 16 18 19 22 131
Pla II 3 5 6.5 8 4 2 12.5 10 1 52
The tabulated value of χ2at 2 d.f. and 5% level of significance is 5.99. Since H
is greater than this value, H0is rejected at 5% level of significance.
e.g.
If the data meet the requirements for a parametric test, it is better to use one -
way independent-measures Analysis of Variance (ANOVA) because it is more
powerful than the Kruskal-Wallis test.
237
The above table indicates that the Khadi textile products are frequently
purchased by the consumers from rural area (mean rank 211.95) compared to
consumers from urban (mean rank 209) and semi -urban areas (mean rank 208.63).
This mean variation is statistically insignificant at 5 per cent level of significance
(P>.05 in above cases)
18.4 REVISION POINTS
1. Sign test
2. One sample sign test
3. Two sample sign test
18.5 INTEXT QUESTIONS
1. List out the advantages and disadvantages of non parametric test
2. Explain the Importance of sign tests
18.6 SUMMARY
After going through this Lesson you should be able to Explain Sign test, one
sample sign test and two sample sign test
18.7 TERMINAL EXERSICE
It turns out that 5 of the participants like the new drink better, and the rest
prefer the old one. At .05 significance level, can we reject the notion that the two
drinks are equally popular?
18.8 SUPPLEMENTARY MATERIALS
Barzun, Jacques, and Graff, Henery, F., The Modern Researcher, rev. ed., New
York: Harcourt, Brace & World, Inc., 1970.
18.9 ASSIGNMENT
The sign test is considered a weaker test, because it tests the pair value below
or above the median and it does not measure the pair difference - Comment
238
239
LESSON – 19
Y1 Y2 Y3
1 X1
1
2 X2 1 11
1 31 3
Y6 6
3 X3 21 2
Y7 7
4 X4 2 2 32
22
5 X5 2
Y4 Y5
4 5
For more information, click on an element of this diagram, or choose from this
list: Latent Constructs Structural Model Structural Error Manifest Variables
Measurement Model Measurement Error
This diagram uses the dominant symbolic language in the SEM world.
However, there are alternate forms, including the "RAM," reticular action model.
Latent Constructs
In structural equation modeling, the key variables of interest are usually
"latent constructs"--abstract psychological concepts such as "intelligence" or
241
"attitude." We can observe the behavior of latent variables only indirectly, and
imperfectly, through their effects on manifest variables.
A structural equation model may include two types of latent
constructs--exogenous and endogenous. In the most
traditional system, exogenous constructs are indicated by
the Greek character "ksi" (at left) and endogenous constructs are
indicated by the Greek character "eta" (at right). These two types of
constructs are distinguished on the basis of whether or not they are dependent
variables in any equation in the system of equations represented by the model.
Exogenous constructs are independent variables in all equations in which they
appear, while endogenous constructs are dependent variables in at least one
equation--although they may be independent variables in other equations in the
system. In graphical terms, each endogenous construct is the target of at least one
one-headed arrow, while exogenous constructs are only targeted by two-headed
arrows.
Structural Model
In SEM, the structural model includes the relationships among the latent
constructs. These relationships are chiefly linear, although flexible extensions to
the basic SEM system allow for the inclusion of nonlinear relations, as well. In the
diagram, one-headed arrows represent regression relationships, while two-headed
arrows represent correlational relations--that is, shared variation that is not
explained within the model.
Parameters representing regression relations between latent
constructs are typically labeled with the Greek character "gamma"
(at left) for the regression of an endogenous construct on an exogenous
construct, or with the Greek character "beta" (at right) for the regression of
one endogenous construct on another endogenous construct. Typically in
SEM, exogenous constructs are allowed to covary freely. Parameters
labeled with the Greek character "phi" (at left) represent these covariances.
This covariance comes from common predictors of the exogenous constructs which
lie outside the model under consideration.
Structural Error
Few SEM researchers expect to perfectly predict their dependent
constructs, so model typically include a structural error term, labeled with
the Greek character "zeta" (at right). To achieve consistent parameter estimation,
these error terms are assumed to be uncorrelated with the model's exogenous
constructs. (Violations of this assumption come about as a result of the excluded
predictor problem.) However, structural error terms may be modeled as being
correlated with other structural error terms. Such a specification indicates that the
endogenous constructs associated with those error terms share common variation
that is not explained by predictor relations in the model.
242
Manifest Variables
SEM researchers use manifest variables--that is, actual
measures and scores--to ground their latent construct models with
real data. Manifest variables associated with exogenous constructs
are labeled X, while those associated with endogenous constructs are labeled Y.
Otherwise, there is no fundamental distinction between these measures, and a
measure that is labeled X in one model may be labeled Y in another.
Measurement Model
In SEM, each latent construct is usually
associated with multiple measures. SEM
researchers most commonly link the latent
constructs to their measures through a factor
analytic measurement model. That is, each
latent construct is modeled as a common
factor underlying the associated measures. These "loadings" linking constructs to
measures are labeled with the Greek character "lambda" (at left). Structural
equation models can include two separate "lambda matrices, one on the X side and
one on the Y side. In SEM applications, the most common measurement model is
the congeneric measurement model, where each measure is associated with only
one latent construct, and all covariation between measures is a consequence of the
relations between measures and constructs.
(Sometimes, however, it makes more sense to model a latent construct as the
result or consequence of its measures. This is the causal indicators model. This
alternative measurement model is also central to Partial Least Squares, a
methodology related to SEM.)
Measurement Error
SEM users typically recognize that their measures are
imperfect, and they attempt to model this imperfection. Thus,
structural equation models include terms representing measurement error. In the
context of the factor analytic measurement model, these measurement error terms
are uniquenesses or unique factors associated wi th each measure. Measurement
error terms associated with X measures are labeled with the Greek character "delta"
(at left) while terms associated with Y measures are labeled with "epsilon" (at right).
Conceptually, almost every measure has an associated error term. In other words,
almost every measure is acknowledged to include some error.
However, when a construct is associated with only a single measure, it is
usually impossible (due to the limits of identification) to estimate the amount of
measurement error within the model. In such cases, the researcher must prespecify
the amount of measurement error before attempting to estimate model parameters.
In this situation, researchers may be tempted to simply assume that there is no
measurement error. However, if this assumption is false, then model parameter
estimates will be biased.
243
by Glymour and others has appeared in the literature on "data mining," a field
which deals with the increasingly common situation of an analyst being
overwhelmed by the quantity of available data.
However, these tools may also be useful in a confirmatory context. Researchers
may wish to evaluate fit information for one model in light of fit information for
other models. If many models fit a data set well, then the good fit of one model
would seem to offer less compelling evidence in favor of the theory that inspired the
model. This procedure might also be sued to find equivalent models empirically.
Spirtes et al. have suggested that augmenting "traditional" SEM analysis with
their exploratory approach is more likely to produce the "true" model for a data set
(if one exists) than is SEM analysis by itself.
In their work, Spirtes et al. use a distinctive, graphically oriented language. A
link between two constructs or variables is called an "edge." If the link represents a
covariance between the two, it is called an "undirected edge"; otherwise, it is called
a "directed edge." Depending on the algorithm used, TETRAD II may begin with a
model where all variables are correlated (an "undirected g raph") and then proceed
to eliminate edges where possible and turn undirected edges into directed edges.
The result is a "partially direct acyclic graph" which fits the data set to a tolerance
specified by the user. The word "acyclic" indicates that the re sulting model will not
involve reciprocal or feedback relations--a limitation of the method.
Tetrad analysis
Latent class analysis (LCA) is appropriate when the dependent variable is
intrinsically categorical--that is, when it boils down to distinguishing between those
who take an action and those who do not, or between those who vote for different
political candidates. In SEM, ordinal measures may be used to reflect continuous
latent variables, but in LCA both the measure and the underlying latent variable
are categorical. For more information, see Keith Markus' LCA Web site, or Karen
Schmidt McCollam's overview chapter in Marcoulides (1998).
Tetrad analysis
Unrestricted factor analysis
(This section consists of edited SEMNET comments from Stan Mulaik.)
Researchers can use unrestricted factor analysis to test only the hypothe sis
that relations between a set of measures can be explained by a certain number of
common factors, without specifying a particular factor structure. This allows the
researcher to determine whether any p-factor model will provide acceptable fit. If
this hypothesis fails, then there is no point in testing more constrained CFA
models.
The unrestricted model was described by Karl Jöreskog in an addendum to a
reprint of the first paper on confirmatory factor analysis that appeared in
Psychometrika in 1972. This addendum is to be found in:
246
modeling there are considerably more parameters. Each fixed parameter in the
model is potentially false and each can take on, in principle, an infinite number of
alternative values. Thus, each fixed parameter needs to be evaluated, in principle,
one at a time.
A method for evaluating the power of the likelihood ratio test in structural
equation modeling was developed by Satorra and Saris (1985). Their method can be
outlined as follows. First, estimate the model of interest. Second, choose a fixed
parameter whose power is desired. Third, re -estimate the initial model with each
estimated parameter fixed at their estimated value and choose an "alternative" fixed
value for the parameter of interest. Note that if the null hypothesis is true for that
parameter, then the likelihood ratio chi -square for the model would be zero with
degrees-of-freedom equaling the degrees-of-freedom of the model. If the null
hypothesis is false for that parameter, then the l ikelihood ratio chi-square will be
some positive number reflecting the specification error incurred by fixing that
parameter to the value chosen in the initial model. This number is the
noncentrality parameter (NCP) of the noncentral chi-square distribution, which is
the distribution of the test statistic when the null hypothesis is false. This number
can be compared to tabled values of the noncentral chi -square distribution to
assess power.
Clearly, the method originally proposed by Satorra and Saris (1985) would be
tedious to carry out in practice. Recen tly, Satorra (1989) recognized that the
modification indices (in LISREL) or the Lagrange multiplier tests (in EQS) are
actually one degree-of-freedom noncentrality parameters. Thus, for any estimated
model, it is a simple matter to look at these indices in relation to tabled value s of
the noncentral chi-square distribution in order to assess power. It should be noted
that power can be assessed for free parameters as well. That is, the square of the T -
value (in LISREL) or the Wald test (in EQS) can be used to assess the power of an
estimated parameter in the model, against a null hypothesis that the value of the
parameter is zero.
Consideration of the power associated with the likelihood ratio test (or other
asymptotically equivalent tests) led to an approach for conducting model
modification. Specifically, Saris, Satorra, and Sörbom (1987) advocated a
combination of the modification index and the expected change statistic (EC) to
guide model modifications. The EC is an estimate of the value that a fixed
parameter will take if that parameter is freed. Thus it represents the "distance"
between the value of the parameter under the null hypothesis (usually zero) and the
value of the parameter under the alternative hypothesis. Therefore, with the
modification index (NCP) and EC in hand one has all the ingredients to engage in
model modification with an eye toward power considerations. This power-based
approach to model modification was advocated by Kaplan (1990, with subsequent
commentary).
250
Under the approach advocated by Kaplan (1990) and Saris, Satorra, and
Sörbom (1987) parameters with large MIs and large ECs should be freed first
provided it makes substantive sense to do so because these parameters are
associated with higher probabilities of being false. Parameters associated with large
MIs and small ECs need to be studied carefully because , as shown by Kaplan
(1989), this might reflect a sample size sensitivity problem. Parameters associated
with small MIs and large ECs also need to be studied carefully in that it may also
reflect a sample size problem. Nevertheless, there may be substantive value in re -
estimating the model with this parameter freed because of the large contribution it
would be expected to make. Finally, parameters associated with small MIs and
small ECs could be ignored. Note that relative comparisons across parameters can
be made by employing standardized or completely standardized ECs (see Kaplan,
1989; Chou and Bentler, 1990).
With the advent of multivariate methods of model modification (see e.g.
Bentler, 1992) the question naturally arises how this might pertain to assessin g
power in more than one parameter simultaneously. Saris, Satorra, and Sörbom
(1987) recognized that power of two parameters of equal "true value" can differ
depending on the location of the parameters in the model. That is, if two
parameters have the same true alternative value, the power associated with each
may be different because of their different statistical associations with other
parameters in the model. The mechanism underlying these associations pertain to
the pattern of zero and non-zero values in the covariance matrix of the parameter
estimates. This matrix and its role in specification error and power have been
discussed recently in Kaplan and Wenger (1993). Suffice to say that multivariate
methods of model modification can, in some specific cases, lead one to miss the
subtle changes that can take place in the underlying system of relationships that
can be observed by a univariate approach to model modification. Thus, while it is
true that using multivariate methods minimizes Type II (or Type I) errors incurred
by model modification relative to univariate approaches, one does so at the expense
of monitoring subtle changes that might impact the substantive interpretation of
the model.
Finally, an issue of some practical importance is the role that power plays in
multiple group structural equation modeling. Recently, Kaplan and George (1995)
studied power in the multiple group confirmatory factor analysis setting.
Specifically, Kaplan and George (1995) examined the power of the Wald test of
factor mean differences under violations of factorial invariance. Using an approach
similar to that used in power studies of Hotelling's T-square, Kaplan and George
(1995) found that power was most affected by the degree of true factor mean
differences. The size of the model was also found to affect the power of the test with
larger models giving rise to increased power. Finally, with equal sample sizes,
Kaplan and George (1995) found that the power of the Wald test of factor mean
differences is relatively robust to violations of factorial invariance. With unequal
251
sample sizes, large changes in the power of the test were observed even under
conditions of factorial invariance.
When framed in terms of decisions errors, the results of Kaplan and George
(1995) suggest that the marginal effect of non -invariance was to decrease the
probability of Type II errors. In contrast, the marginal effect of inequality of sample
size was to increase Type II errors. Kaplan and George (1995) concluded with
practical guidelines suggesting that if the hypothesis of factor mean differences was
rejected under conditions of unequal sample size, then it was probably the case
that the null hypothesis was false. On the other hand, if the hypothesis was not
rejected, then it could be the effect of unequal sample siz es, lack of substantively
large differences, or both.
19.4 REVISION POINTS
1. Structural Equation Modelling
2. Form of Structural Equation Modelling
3. Latent construct
4. Structural model
5. Structural error
6. Manifest Variables
7. CFA
19.5 INTEXT QUESTIONS
1. Write short notes on
a) Latent Construct
b) Structural Model
c) Structural error
d) Manifest Variables
e) CFA
19.6 SUMMARY
After going through this Lesson you should be able to Explain Structural
Equation Modelling, form of Structural Equation Modelling, late nt construct,
Structural model, structural error, Manifest Variables, CFA and PLS
19.7 TERMINAL EXERSICE
1. Kindly explain the importance of structural equation modeling
19.8 SUPPLEMENTARY MATERIALS
Saris, W. E., & Satorra, A. (1987). Characteristics of structural equation
models which affect the power of the likelihood ratio test. In W.E. Saris & I. N.
Gallhofer (Eds.), Sociometric research (Vol. 2). London: Macmillan.
19.9 ASSIGNMENT
1. Briefly explain when to use structural equation modeling. Discuss with
examples.
252
253
LESSON – 20
CLUSTER ANALYSIS
20.1 INTRODUCTION
The statistical procedure used to form groups that share similar
characteristics is called cluster analysis. The choice of variables is crucial to cluster
analysis since only those used will determine the clusters or groups. The concepts
of DISTANCE and SIMILARITY are key ingredients in this statistical procedure.
Distance is the measure of how far apart cases are, whereas, similarity measures
the closeness of cases within a specific group or cluster. Distance measures are
SMALL while similarity measures are LARGE. Within cluster analysis, cases are
grouped together on the basis of their “nearness”. Cluster Analysis usually employs
the DISTANCE measure (how far apart cases are) in defining the clusters or groups.
20.2 OBJECTIVE
After going through this Lesson you should be able to understand Cluster
analysis, K-Mean Cluster analysis, Multi dimensional scaling and logic of Multi
dimensional scaling
20.3 CONTENT
20.3.1 Cluster analysis,
20.3.2 K-Mean Cluster analysis,
20.3.3 Multi dimensional scaling
20.3.4 Logic of Multi dimensional scaling
20.3.1 Cluster Analysis
Cluster analysis technique is employed in this study to segment the
respondents in to two or more similar groups. The factors derived from the Factor
Analyses in relation to Shopping Behaviour and Purchase Decision and the
satisfaction scores were used for this analysis. The main purpose of this
classification is to see whether respondents can be grouped into similar patterns.
These groups will be further used to analyze whether there is association between
respondents of different segments based on personal variables of respondents. At
this juncture K- means cluster analysis was applied to classify the consumer’s
perception and purchase decision. Hierarchical cluster method with agglomeration
schedule decides the number of clusters. It was found that three large coefficients
in the agglomeration schedule were significantly large in their differences.
Therefore, it was concluded that there was an existence of three clusters which was
presented as follows:
20.3.2 K-Means Cluster Analysis
There are different classification techniques to perform cluster analysis. K-
means cluster attempts to identify relatively homogeneous groups of cases based on
selected characteristics, using an algorithm that can handle large numbers of
cases. However, the algorithm requires specifying the number of clusters. Number
of initial cluster centers can be specified before if this information is known.
254
From the above description we find that clustering methods in general are
judgemental and are devoid of statistical inferences. For problems concerning large
number of variables, various cut-andtry methods have been proposed for locating
clusters. McQuitty has specially developed a number of rather elaborate
computational routines* for that purpose.
In spite of the above stated limitation, cluster an alysis has been found useful
in context of market research studies. Through the use of this technique we can
make segments of market of a product on the basis of several characteristics of the
customers such as personality, socio-economic considerations, psychological
factors, purchasing habits and like ones.
8.3.4 Multidimensional Scaling
Multidimensional scaling (MDS) allows a researcher to measure an item in
more than one dimension at a time. The basic assumption is that people perceive a
set of objects as being more or less similar to one another on a number of
dimensions (usually uncorrelated wi th one another) instead of only one.
There are several MDS techniques (also known as techniques for dimensional
reduction) often used for the purpose of revealing patterns of one sort or another in
interdependent data structures. If data happen to be non-metric, MDS involves
rank ordering each pair of objects in terms of similarity. Then the judged
similarities are transformed into distances through statistical manipulations and
are consequently shown in n-dimensional space in a way that the interpoint
distances best preserve the original interpoint proximities. After this sort of
mapping is performed, the dimensions are usually interpreted and labeled by the
researcher.
The significance of MDS lies in the fact that it enables the rese archer to study
“The perceptual structure of a set of stimuli and the cognitive processes underlying
the development of this structure.... MDS provides a mechanism for determining
the truly salient attributes without forcing the judge to appear irrational.” With
MDS, one can scale objects, individuals or both with a minimum of information.
The MDS analysis will reveal the most salient attributes which happen to be the
primary determinants for making a specific decision.
Multidimensional scaling (MDS) can be considered to be an alternative to factor
analysis (see Factor Analysis). In general, the goal of the analysis is to detect
meaningful underlying dimensions that allow the researcher to explain observed
similarities or dissimilarities (distances) between the investigated objects. In factor
analysis, the similarities between objects (e.g., variables) are expressed in the
correlation matrix. With MDS, you can analyze any kind of similarity or
dissimilarity matrix, in addition to correlation matrices.
8.3.5 Logic of MDS
The following simple example may demonstrate the logic of an MDS analysis.
Suppose we take a matrix of distances between major US cities from a map. We
then analyze this matrix, specifying that we want to reproduce the distances based
256
on two dimensions. As a result of the MDS analysis, we would most likely obtain a
two-dimensional representation of the locations of the cities, that is, we would
basically obtain a two-dimensional map.
In general then, MDS attempts to arrange "objects" (major cities in this
example) in a space with a particular number of dimensions (two-dimensional in
this example) so as to reproduce the observed distances. As a result, we can
"explain" the distances in terms of underlying dimensions; in our example, we could
explain the distances in terms of the two geographical dimensions: north/south
and east/west.
Orientation of axes: As in factor analysis, the actual orientation of axes in the
final solution is arbitrary. To return to our e xample, we could rotate the map in any
way we want, the distances between cities remain the same. Thus, the final
orientation of axes in the plane or space is mostly the result of a subjective decision
by the researcher, who will choose an orientation that can be most easily explained.
To return to our example, we could have chosen an orientation of axes other than
north/south and east/west; however, that orientation is most convenient because it
"makes the most sense" (i.e., it is easily interpretable).
Computational Approach
MDS is not so much an exact procedure as rather a way to "rearrange" objects
in an efficient manner, so as to arrive at a configuration that best approximates the
observed distances. It actually moves objects around in the space defi ned by the
requested number of dimensions, and checks how well the distances between
objects can be reproduced by the new configuration. In more technical terms, it
uses a function minimization algorithm that evaluates different configurations with
the goal of maximizing the goodness-of-fit (or minimizing "lack of fit").
Measures of goodness-of-fit: Stress. The most common measure that is used
to evaluate how well (or poorly) a particular configuration reproduces the observed
distance matrix is the stress measure. The raw stress value Phi of a configuration is
defined by:
Shepard diagram. You can plot the reproduced distances for a particular
number of dimensions against the observed input data (distances). This scatterplot
is referred to as a Shepard diagram. This plot shows the reproduced distances
plotted on the vertical (Y) axis versus the original similarities plotted on the
horizontal (X) axis (hence, the generally negative slope). This plot also shows a step-
function. This line represents the so- called D-hat values, that is, the result of the
monotone transformation f( ) of the input data. If all reproduced distances fall
onto the step-line, then the rank-ordering of distances (or similarities) would be
perfectly reproduced by the respective solution (dimensional model). Deviations
from the step-line indicate lack of fit.
How Many Dimensions to Specify?
If you are familiar with factor analysis, you will be quite aware of this issue. If
you are not familiar with factor analysis, you may want to read the Factor Analysis
section in the manual; however, this is not necessary in order to understand the
following discussion. In general, the more dimensions we use in order to reproduce
the distance matrix, the better is the fit of the reproduced matrix to the observed
matrix (i.e., the smaller is the stress). In fact, if we use as many dimensions as
there are variables, then we can perfectly reproduce the observed distance matrix.
Of course, our goal is to reduce the observed complexity of nature, that is, to
explain the distance matrix in terms of fewer underlying dimensions. To return to
the example of distances between cities, once we have a two-dimensional map it is
much easier to visualize the location of and navigate between cities, as compared to
relying on the distance matrix only.
Sources of misfit. Let's consider for a moment why fewer factors may produce
a worse representation of a distance matrix than would more factors. Imagine the
three cities A, B, and C, and the three cities D, E, and F; shown below are their
distances from each other.
A B C D E F
0
0
A 9 D
0 90 0
B 0 E
9 18 9
C 9 0 F 0
0 0 0
0
In the first matrix, all cities are exactly 90 miles apart from each other; in the
second matrix, cities D and F are 180 miles apart. Now, can we arrange the three
cities (objects) on one dimension (line)? Indeed, we can arrange cities D, E, and F
on one dimension:
D---90 miles---E---90 miles---F
D is 90 miles away from E, and E is 90 miles away from F; thus, D is
90+90=180 miles away from F. If you try to do the same thing with cities A, B, and
C you will see that there is no way to arrange the three cities on one line so that the
258
90 90
miles miles
B 90 miles C
Arranging the three cities in this manner, we can perfectly reproduce the
distances between them. Without going into much detail, this small example
illustrates how a particular distance matrix implies a particular number of
dimensions. Of course, "real" data are never this "clean," and contain a lot of noise,
that is, random variability that contributes to the differences between the
reproduced and observed matrix.
Scree test: A common way to decide how many dimensions to use is to plot
the stress value against different numbers of dimensions. This test was first
proposed by Cattell (1966) in the context of the number-of-factors problem in factor
analysis (see Factor Analysis); Kruskal and Wish (1978; pp. 53-60) discuss the
application of this plot to MDS.
Cattell suggests to find the place where the smooth decrease of stress values
(eigenvalues in factor analysis) appears to level off to the right of the plot. To the
right of this point, you find, presumably, only "factorial scree" - "scree" is the
geological term referring to the debris which collects on the lower part of a rocky
slope.
Interpretability of Configuration: A second criterion for deciding how many
dimensions to interpret is the clarity of the final configuration. Sometimes, as in
our example of distances between cities, the resultant dimensions are easily
interpreted. At other times, the points in the plot form a sort of "random cloud," and
there is no straightforward and easy way to interpret the dimensions. In the latter
case, you should try to include more or fewer dimensions and examine the
resultant final configurations. Often, more interpretable solutions emerge. However,
if the data points in the plot do not follow any pattern, and if the stress plot does
not show any clear "elbow," then the data are most likely random "noise."
Interpreting the Dimensions
The interpretation of dimensions usually represents the final step of the
analysis. As mentioned earlier, the actual orientations of the axes from the MDS
analysis are arbitrary, and can be rotated in any direction. A first step is to produce
scatterplots of the objects in the different two-dimensional planes.
In addition to "meaningful dimensions," you should also look for clusters of
points or particular patterns and configurations (such as circles, manifolds, etc.).
For a detailed discussion of how to interpret final configurations, see Borg and
Lingoes (1987), Borg and Shye (in press), or Guttman (1968).
259
calculations, which can now be left to PCs using accessible and user-friendly
Statistical JavaScript Applets.
Today's good decisions are driven by data. In all aspects of our lives, and
importantly in the business context, an amazing diversity of data is available for
inspection and enlightenment. Moreover, business managers and professionals are
increasingly encouraged to justify decisions on the basis of data.
Business managers need statistical model-based decision support systems.
Statistical skills enable you to intelligently collect, analyze and interpret data
relevant to their decision-making. Statistical concepts and statistical thinking
enable you:
to solve problems in a diversity of contexts
to add substance to decisions.
This Web site is a course in statistics appreciation; i.e., to acquire a feel for the
statistical way of thinking. Appreciation of statistics is wonderful: it makes what is
excellent in statistical thinking belong to you as well. It is an introductory course in
statistics that is designed to provide you with the basic concepts, and methods of
statistical analysis for processes and products. Materials in this Web site are
tailored to meet your needs in making good decision and they promote one to think
statistically. The cardinal objective for this Web site is to increase the extent to
which statistical thinking is embedded with managerial thinking for decision
making under uncertainties.
In competitive environment, business managers must design quality into
products, and into the processes of making the products. They must facilitate a
process of never-ending improvement at all stages of manufacturing and service.
This is a strategy that employs statistical methods, particularly statistically
designed experiments, and produces processes that provide high yield an d products
that seldom fail. Moreover, it facilitates development of robust products that are
insensitive to changes in the environment and internal component variation.
Carefully planned statistical studies remove hindrances to high quality and
productivity at every stage of production. This saves time and money. It is well
recognized that quality must be engineered into products as early as possible in the
design process. One must know how to use carefully planned, cost-effective
statistical experiments to improve, optimize and make robust products and
processes.
20.4 REVISION POINTS
1. Cluster analysis,
2. K-Mean Cluster analysis,
3. Multi dimensional scaling
4. Logic of Multi dimensional scaling
261
Satorra, A., & Saris, W.E. (1985). Power of the likelihood ratio test in
covariance structure analysis. Psychometrika, 50, 83-90.
20.11 LEARNING ACTIVITY
Explain briefly about the advantages of Cluster analysis. Enlist the limitation
of Cluster analysis
20.12 KEYWORD
Cluster analysis, K-Mean Cluster analysis, Multi-dimensional scaling and logic
of Multi-dimensional scaling
263
UNIT – VI
LESSON – 21
RESEARCH REPORT
21.1 INTRODUCTION
In this chapter deals with Research Report - Structure of Research Report – Title
of a Report – Table of contents - Synopsis – Introduction and and body of a report -
Table and diagrams – Results and Discussions – Recommendations, conclusion, and
scope for further research-reference –Appendices.
There are people who do not consider writing of report as an integral part ofthe
research process. But the general opinion is in favour of treating the presentation of
research results or the writing of report as part and parcel of the research project.
Writing of report is the last step in a research study and requires a set of skills
somewhat different from those called for in respect of the earlier stages of research.
This task should be accomplished by the researcher with utmost care; he may seek
the assistance and guidance of experts for the purpose.
21.2 OBJECTIVE
After going through this Lesson you should be able to understand what is
report, report writing Steps in report writing and lay out of research report
21.3 CONTENT
9.3.1 WHAt is report,
9.3.2 Report writing
9.3.3 Steps in report writing and
9.3.4 Lay out of research report
DIFFERENT STEPS IN WRITING REPORT
Research reports are the product of slow, painstaking, accurate inductive
work. The usual steps involved in writing report are:
(a) logical analysis of the subject-matter; (b) preparation of the final outline;
(c) preparation of the rough draft; (d) rewriting and polishing; (c) preparation of the
final bibliography; and (f) writing the final draft. Though all these steps are self
explanatory, yet a brief mention of each one of these will be appropriate for better
understanding.
a) Logical analysis of the subject matter
It is the first step which is primarily concerned with the development of a
subject. There are two ways in which to develop a subject (a) logically and (b)
chronologically. The logical development is made on the basis of mental
connections and associations between the one thing and another by means of
analysis. Logical treatment often consists in developing the material from the
simple possible to the most complex structures. Chronological development is
264
4. Number of volumes.
Example
Kothari, C.R., Quantitative Techniques, New Delhi, Vikas Publishing House
Pvt. Ltd., 1978.
For magazines and newspapers the order may be as under:
1. Name of the author, last name first.
2. Title of article, in quotation marks.
3. Name of periodical, underlined to indicate italics.
4. The volume or volume and number.
5. The date of the issue.
6. The pagination.
Example
Robert V. Roosa, “Coping with Short-term International Money Flows”, The
Banker, London, September, 1971, p. 995.
The above examples are just the samples for bibliography entries and may be
used, but oneshould also remember that they are not the only acceptable forms.
The only thing important is that, whatever method one selects, it must remain
consistent.
f) Writing the final draft
This constitutes the last step. The final draft should be written in a concise
and objective style and in simple language, avoiding vague expressions such as “it
seems”, “there may be”, and the like ones. While writing the final draft, the
researcher must avoid abstract terminology and technical jargon. Illustrations and
examples based on common experiences must be incorporated in the final draft as
they happen to be most effective in communicating the research findings to others.
A research report should not be dull, but must enthuse people and maintain
interest and must show originality. It must be remembered that every report should
be an attempt to solve some intellectual problem and must contribute to the
solution of a problem and must add to the knowledge of both the researcher and
the reader.
LAYOUT OF THE RESEARCH REPORT
Anybody, who is reading the research report, must necessarily be conveyed
enough about the study so that he can place it in its general scientific context,
judge the adequacy of its methods and thus form an opinion of how seriously the
findings are to be taken. For this purpose there is the need of proper layout of the
report. The layout of the report means as to what the research report should
contain. A comprehensive layout of the research report should comprise (A)
preliminary pages; (B) the main text; and (C) the end matter. Let us deal with them
separately.
266
a) Preliminary Pages
In its preliminary pages the report should carry a title and date, followed by
acknowledgements in the form of ‘Preface’ or ‘Foreword’. Then there should be a
table of contents followed by list of tables and illustrations so that the decision -
maker or anybody interested in reading the report can easily locate the required
information in the report.
b) Main Text
The main text provides the complete outline of the research report along with
all details. Title of theresearch study is repeated at the top of the first page of the
main text and then follows the other details on pages numbered consecutively,
beginning with the second page. Each main section of the report should begin on a
new page. The main text of the report should have the following sections: (i)
Introduction; (ii) Statement of findings and recommendations; (iii) The results;
(iv)The implications drawn from the results; and (v) The summary.
i) Introduction
The purpose of introduction is to introduce the research project to the readers.
It should contain a clear statement of the objectives of research i.e., enough
background should be given to make clear to the reader why the problem was
considered worth investigating. A brief summary of other relevant research may
also be stated so that the present study can be seen in that context. The
hypotheses of study, if any, and the definitions of the major concepts employed in
the study should be explicitly stated in the introduction of the report. The
methodology adopted in conducting the study must be fully explained. The
scientific reader would like to know in detail about such thing: How was the study
carried out? What was its basic design? If the study was an experimental one, then
what were the experimental manipulations? If the data were collected by means of
questionnaires or interviews, then exactly what questions were asked (The
questionnaire or interview schedule is usually given in an appendix)? If
measurements were based on observation, then what instructions were given to the
observers? Regarding the sample used in the study the reader should be told: Who
were the subjects? How many were there? How were they selected? All these
questions are crucial for estimating the probable limits of generalizability of the
findings. The statistical analysis adopted must also be clearly stated. In addition to
all this, the scope of the study should be stated and the boundary lines be
demarcated. The various limitations, under which the research project was
completed, must also be narrated.
ii) Statement of findings and recommendations
After introduction, the research report must contain a statement of findings
and recommendations in non-technical language so that it can be easily
understood by all concerned. If the findings happen to be extensive, at this point
they should be put in the summarized form.
267
iii) Results
A detailed presentation of the findings of the study, with supporting data in
the form of tables and charts together with a validation of results, is the next step
in writing the main text of the report. This generally comprises the main body of the
report, extending over several chapters. The result section of the report should
contain statistical summaries and reductions of the data rather than the raw data.
All the results should be presented in logical sequence and splitted into readily
identifiable sections. All relevant results must find a place in the report. But how
one is to decide about what is relevan t is the basic question. Quite often guidance
comes primarily from the research problem and from the hypotheses, if any, with
which the study was concerned. But ultimately the researcher must rely on his own
judgement in deciding the outline of his report. “Nevertheless, it is still necessary
thathe states clearly the problem with which he was concerned, the procedure by
which he worked onthe problem, the conclusions at which he arrived, and the
bases for his conclusions.”
iv) Implications of the results
Toward the end of the main text, the researcher should again put down the
results of his research clearly and precisely. He should, state the implications that
flow from the results of the study, for the general reader is interested in the
implications for understanding the human behaviour. Such implications may have
three aspects as stated below:
(a) A statement of the inferences drawn from the present study which may be
expected to apply in similar circumstances.
(b) The conditions of the present study whi ch may limit the extent of legitimate
generalizations of the inferences drawn from the study.
(c) The relevant questions that still remain unanswered or new questions
raised by the study along with suggestions for the kind of research that would
provide answers for them.
It is considered a good practice to finish the report with a short conclusion
which summarises andrecapitulates the main points of the study. The conclusion
drawn from the study should be clearly related to the hypotheses that were stated
in the introductory section. At the same time, a forecast of the probable future of
the subject and an indication of the kind of research which needs to be done in that
particular field is useful and desirable.
v) Summary
It has become customary to conclude the research report with a very brief
summary, resting in brief the research problem, the methodology, the major
findings and the major conclusions drawn from the research results.
c) End Matter
At the end of the report, appendices should be enlisted in respect of all
technical data such as questionnaires, sample information, mathematical
derivations and the like ones. Bibliography of sources consulted should also be
given. Index (an alphabetical listing of names, places and topics along with the
268
LESSON – 22
TYPES OF REPORT
22.1 INTRODUCTION
Research reports vary greatly in length and type. In each individual case, both
the length and theform are largely dictated by the problems at hand. For instance,
business firms prefer reports in theletter form, just one or two pages in length.
Banks, insurance organisations and financial institutionsare generally fond of the
short balance-sheet type of tabulation for their annual reports to theircustomers
and shareholders. Mathematicians prefer to write the results of their investigations
in theform of algebraic notations. Chemists report their results in symbols and
formulae. Students ofliterature usually write long reports presenting the critical
analysis of some writer or period or the likewith a liberal use of quotations from the
works of the author under discussion. In the field of educationand psychology, the
favourite form is the report on the results of experimentation accompanied bythe
detailed statistical tabulations. Clinic al psychologists and social pathologists
frequently find itnecessary to make use of the case -history form.
22.2 OBJECTIVE
After going through this Lesson you should be able to understand what are the
Types of report, Technical report, Popular report and Oral presentation
22.3 CONTENT
22.3.1 Types of report,
22.3.2 Technical report,
22.3.3 Popular report and
22.3.4 Oral presentation
22.3.1 Types of Report
News items in the daily papers are also forms of report writing. They represent
firsthand on-the-scene accounts of the events described or compilations of
interviews with persons who were on thescene. In such reports the first paragraph
usually contains the important information in detail and thesucceeding paragraphs
contain material which is progressively less and less important.
Book-reviews which analyze the content of the book and report on the author’s
intentions, hissuccess or failure in achieving his aims, his language, his style,
scholarship, bias or his point of view.Such reviews also happen to be a kind of
short report. The reports prepared by governmental bureaus,special commissions,
and similar other organisations are generally very comprehensive reports onthe
issues involved. Such reports are usually considered as important research
products. Similarly,Ph.D. theses and dissertations are also a form of report-writing,
usually completed by students inacademic institutions.
The above narration throws light on the fact that the results of a research
investigation can bepresented in a number of ways viz., a technical report, a
popular report, an article, a monograph or attimes even in the form of oral
270
LESSON – 23
Example
John Gassner, Masters of the Drama, New York: Dover Publications, Inc.
1954, p. 315.
ii) Regarding multivolumed reference
1. Author’s name in the normal order;
2. Title of work, underlined to indicate italics;
3. Place and date of publication;
4. Number of volume;
5. Pagination references (The page number).
iii) Regarding works arranged alphabetically
For works arranged alphabetically such as encyclopedias and dictionaries, no
pagination reference is usually needed. In such cases the order is illustrated as
under:
Example 1
“Salamanca,” Encyclopaedia Britannica, 14th Edition.
Example 2
“Mary Wollstonecraft Godwin,” Dictionary of national biography.
But if there should be a detailed reference to a long encyclopedia article,
volume and pagination reference may be found necessary.
iv) Regarding periodicals reference
1. Name of the author in normal order
2. Title of article, in quotation marks
3. Name of periodical, underlined to indicate italics
4. Volume number
5. Date of issuance
6. Pagination
v) Regarding anthologies and collections, reference
Quotations from anthologies or collections of literary works must be
acknowledged not only by author, but also by the name of the collector.
vi) Regarding second-hand quotations, reference
In such cases the documentation should be handled as follows:
1. Original author and title;
2. “quoted or cited in,”;
3. Second author and work.
Example
J.F. Jones, Life in Ploynesia, p. 16, quoted in History of the Pacific Ocean area,
by R.B. Abel, p. 191.
277
bk., book
bull., bulletin
cf., compare
ch., chapter
col., column
diss., dissertation
ed., editor, edition, edited.
ed. cit., edition cited
e.g., exempli gratia: for example
eng., enlarged
et.al., and others
et seq., et sequens: and the following
ex., example
f., ff., and the following
fig(s)., figure(s)
fn., footnote
ibid., ibidem: in the same place (when two or more successive footnotes refer
to the same work, it is not necessary to repeat complete reference for the second
footnote. Ibid. may be used. If different pages are referred to, pagination must be
shown).
id., idem: the same
ill., illus., orillust(s) illustrated, illustration(s)
Intro., intro., introduction
l, or ll, line(s)
loc. cit., in the place cited; used as op.cit., (when new reference
loco citato: is made to the same pagination as cited in the previous note)
MS., MSS., Manuscript or Manuscripts
N.B., nota bene: note well
n.d., no date
n.p., no place
no pub., no publisher
no(s)., number(s)
o.p., out of print
op. cit: in the work cited (If reference has been made to a work
279
opera citato and new reference is to be made, ibid., may be used, if intervening
reference has been made to different works, op.cit. must be used. Thename of the
author must precede.
p. or pp., page(s)
passim: here and there
post: after
rev., revised
tr., trans., translator, translated, translation
vid or vide: see, refer to
viz., namely
vol. or vol(s)., volume(s)
vs., versus: against
8. Use of statistics, charts and graphs
A judicious use of statistics in research reports is oftenconsidered a virtue for
it contributes a great deal towards the clarification and simplification of the
material and research results. One may well remember that a good picture is often
worth more than a thousand words. Statistics are usually presented in the form of
tables, charts, bars and line-graphsand pictograms. Such presentation should be
self explanatory and complete in itself. It should be suitable and appropriate
looking to the problem at hand. Finally, statistical presentation should be neatand
attractive.
9. The final draft
Revising and rewriting the rough draft of the report should be done with great
care before writing the final draft. For the purpose, the researcher should put to
himself questionslike: Are the sentences written in the report clear? Are they
grammatically correct? Do they say what is meant’? Do the various points
incorporated in the report fit together logically? “Having atleast one colleague read
the report just before the final revision is extremely helpful. Sentences that seem
crystal-clear to the writer may prove quite confusing to other people; a connection
that had seemed self evident may strike others as a non -sequitur. A friendly critic,
by pointing out passages that seem unclear or illogical, and perhaps suggesting
ways of remedying the difficulties, can be an invaluable aid in achieving the goal of
adequate communication.”
10. Bibliography
Bibliography should be prepared and appended to the research report as
discussed earlier.
11. Preparation of the index
At the end of the report, an index should invariably be given, the value of
which lies in the fact that it acts as a good guide, to the reader. Index may be
prepared both as subject index and as author index. The former gives the names of
280
the subject-topics or conceptsalong with the number of page s on which they have
appeared or discussed in the report, where as the latter gives the similar
information regarding the names of authors. The index should always be arranged
alphabetically. Some people prefer to prepare only one index common for names of
authors, subject-topics, concepts and the like ones.
23.4 REVISION POINTS
1. Size and Physical design
2. Procedure
3. Lay out
4. Treatment of quotation
5. Foot note
6. Documentation style
7. Punctuations
8. Use of Statistical analysis
9. Final draft
10. Bibliography
11. Index
23.5 INTEXT QUESTIONS
Briefly explain the role of statistics in a research report?
23.6 SUMMARY
After going through this Lesson you should be able to explain what are the
Mechanics of writing a report
23.7 TERMINAL EXERSICE
Explain the importance of Lay out while preparing a research report
23.8 SUPPLEMENTARY MATERIALS
Roscoe, John T., Fundamental Research Statistics for the Behavioral
Sciences, New York: Holt, Rinehart and Winston, Inc., 1969.
23.9 ASSIGNMENT
Explain the importance of Size and physical design in the mechanics of the
report?
23.10 SUGGESTED READING REFERENCE BOOKS
Redman, L.V., and Mory, A.V.H., The Romance of Research, 1923.
Roscoe, John T., Fundamental Research Statistics for the Behavioral
Sciences, New York: Holt, Rinehart and Winston, Inc., 1969.
Runyon, Richard P., Inferential Statistics, Philippines: Addison -Wesley
Publishing Company, Inc., 1977.
Sadhu, A.N., and Singh, Amarjit, Research Methodology in Social
Sciences,Bombay: Himalaya Publishing House, 1980.
281
Seboyar, G.E., Manual for Report and Thesis Writing, New York: F.S. Crofts
& Co., 1929.
Selltiz, Claire: Jahoda, Marie, Deutsch, Morton, and Cook, Stuart W.,
Research Methods in Social Relations, rev. ed. New York: Holt, Rinehart
and Winston, Inc., 1959.
Sharma, B.A.V., et al., Research Methods in Social Sciences, New Delhi:
Sterling Publishers Pvt. Ltd., 1983.
Sharma, H.D., and Mukherji, S.P., Research in Economics and Commerce,
Methodology and Surveys, Varanasi: Indian Biographic Centre, 1976.
23.11 LEARNING ACTIVITY
Briefly Explain the Mechanics of a good Report?
23.12 KEYWORD
Size and Physical design, Procedure, Lay out, Treatment of quotation, Foot
note, Documentation style, Punctuations, Use of Statistical analysis, Final draft,
Bibliography, Index
282
LESSON – 24
18. Bowley, A.L., Elements of Statistics, 6th ed. London: P.S. King and Staples
Ltd., 1937.
19. Burgess, Ernest W., “Research Methods in Sociology” in Georges Gurvitch
and W.E. Moore (Ed.), Twentieth Century Sociology, New York: New York
Philosophical Library, 1949.
20. Chance, William A., Statistical Methods for Decision Making, Bombay:
D.B. Taraporevala Sons & Co. Pvt. Ltd., 1975.
21. Chaturvedi, J.C., Mathematical Statistics, Agra: Nok Jhonk Karyalaya,
1953.
22. Chou, Ya-Lun, Statistical Analysis with Business and Economic
Applications, 2nd ed. New York: Holt, Rinehart & Winston, 1974.
23. Clover, Vernon T., and Balsley, Howard L., Business Research Methods,
Columbus, O.: Grid, Inc., 1974.
24. Cochran, W.G., Sampling Techniques, 2nd ed. New York: John Wiley &
Sons., 1963.
25. Cooley, William W., and Lohnes, Paul R., Multivariate Data A nalysis, New
York: John Wiley & Sons.,1971.
26. Croxton, F.E., Cowden, D.J., and Klein, S., Applied General Statistics, 3rd
ed., New Delhi: Prentice-Hall ofIndia Pvt. Ltd., 1975.
27. Dass, S.L., Personality Assessment Through Projective Movie Pictures, New
Delhi: S. Chand & Co. (Pvt.) Ltd., 1974.
28. Davis, G.B., “Introduction to Computers,” 3rd ed., McGraw-Hill
International Book Co., 1981.
29. Deming, W. Edwards., Sample Design in Business Research, New York:
John Wiley & Sons., Inc., 1960.
30. Dennis, Child, The Essentials of Factor Analysis, New York: Holt, Rinehart
and Winston, 1973.
31. Denzin, Norman, The Research Act, Chicago: Aldine, 1973.
32. Edwards, Allen, Statistical Methods, 2nd ed., New York: Holt, Rinehart &
Winston, 1967.
33. Edwards, Allen L., Techniques of Attitude Scale Construction, New York:
Appleton-Century-Crofts, 1957.
34. Emory, C. William, Business Research Methods, Illinois: Richard D. Irwin,
Inc. Homewood, 1976.
35. Ferber, Robert (ed.), Handbook of Marketing Research, New York: McGraw-
Hill, Inc., 1948.
36. Ferber, R., and Verdoorn, P.J., Research Methods in Economics and
Business, New York: The Macmillan Company, 1962.
287
37. Ferguson, George A., Statistical Analysis in Psychology and Education, 4th
ed., New York: McGrawHill Book Co., Inc., 1959.
38. Festinger, Leon and Katz, Daniel (Eds.), Research Methods in the
Behavioral Sciences, New Delhi: Amerind Publishing Co. Pvt. Ltd., Fourth Indian
Reprint, 1976.
39. Fiebleman, J.K., Scientific Method, Netherlands: Martinus Nijhoff, The
Hague, 1972.
40. Fisher, R.A., Statistical Methods for Research Workers, 13th ed., New
York: Hafner Publishing Co., 1958.
41. Fisher, R.A., The Design of Experiments, 7th rev. ed., New York: Hafner
Publishing Co., 1960.
42. Fox, James Harold, Criteria of Good Research, Phi Delta Kappa, Vol. 39
(March 1958).
43. Freedman, P., The Principles of Scientific Research, 2nd ed., New York:
Pergamon Press, 1960.
44. Fruchter, Benjamin, Introduction to Factor Analysis, Princeton, N.J.:
D.Van Nostrand, 1954.
45. Gatner, Elliot S.M., and Cordasco, Francesco, Research and Report
Writing, New York: Barnes & Noble, Inc., 1956.
46. Gaum, Carl G., Graves, Harod F., and Hoffman, Lyne, S.S., Report Writing,
3rd ed., New York: Prentice-Hall, 1950.
47. Ghosh, B.N., Scientific Methods and Social Research, New Delhi: Sterling
Publishers Pvt. Ltd., 1982.
48. Gibbons, J.D., Nonparametric Statistical Inference, Tokyo: McGraw-Hill
Kogakusha Ltd., (International Student Edition), 1971.
49. Giles, G.B., Marketing, 2nd ed., London: Macdonal d & Evans Ltd., 1974.
50. Glock, Charles Y., Survey Research in the Social Sciences, New York:
Russell Sage Foundation, 1967.
51. Godfrey, Arthur, Quantitative Methods for Managers, London: Edward
Arnold (Publishers) Ltd., 1977.
52. Good, Carter V., and Douglas, E. Scates, Methods of Research—
Educational, Psychological, Sociological, New York: Appleton-Century-Crofts, Inc.,
1954.
53. Goode, William J., and Hatt, Paul K., Methods in Social Research, New
York: McGraw-Hill, 1952.
54. Gopal, M.H., An Introduction to Research Procedure in Social
Sciences,Bombay: Asia Publishing House, 1964.
55. Gopal, M.H., Research Reporting in Social Sciences, Dharwar: Karnatak
University, 1965.
288
56. Gorden, Raymond L., Interviewing: Strategy, Techniques and Tactics, rev.
ed., Homewood, Ill.: Dorsey Press, 1975.
57. Green, Paul E., Analyzing Multivariate Data, Hinsdale, Ill.: Dryden Press,
1978.
58. Green, Paul E., and Carmone, F.J., Multidimensional Scaling in Marketing
Analysis, Boston: Allyn&Bacon, Inc., 1970.
59. Guilford, J.P., Psychometric Methods, New York: McGraw Hill, Inc., 1954.
60. Harnett, Donald L., and Murphy, James L., Introductory Statistical
Analysis,Philippines: AddisonWesley Publishing Co., Inc., 1975.
61. Hillway, T.,Introduction to Research, 2nd ed., Boston: Houghton Mifflin,
1964.
62. Hollander, Myles, and Wolfe, Douglas A., Nonparametric Statistical
Methods, New York: John Wiley, 1973.
63. Hunt, R., and Shelley, J., “Computers and Common Sense,” 3rd ed., New
Delhi: Prentice-Hall of India Ltd., 1984.
64. Hyman, Herbert H., et al., Interviewing in Social Research, Chicago:
University of Chicago Press, 1975.
65. John, Peter W.M., Statistical Design and Analysis of Experiments, New
York: The Macmillan Co., 1971.
66. Johnson, Ellen, The Research Report: A Guide for the Beginner, New York:
Ronald Press, 1951.
67. Johnson, Rodney D., and Siskin, Bernard R., Quantitative Techniques for
Business Decisions, New Delhi: Prentice-Hall of India Pvt. Ltd., 1977.
68. Kahn, Robert L. and Cannell, Charles F., The Dynamics of In terviewing,
New York: John Wiley & Sons, 1957.
69. Karson, Marvin J., Multivariate Statistical Methods, Ames, Iowa: The Iowa
State University Press, 1982.
70. Kendall, M.G., A Course in Multivariate Analysis, London, Griffin, 1961.
71. Kerlinger, Fred N. and Pedhazur, Elazar J., Multiple Regression in
Behavioral Research, New York: Holt, Rinehart and Winston, 1973.
72. Kerlinger, Fred N., Foundations of Behavioral Research, 2nd ed., New
York: Holt, Reinhart and Winston, 1973.
73. Kish, Leslie., Survey Sampling, New York: John Wiley & Sons, Inc., 1965.
74. Kothari, C.R., Quantitative Techniques, 2nd ed., New Delhi: Vikas
Publishing House Pvt. Ltd., 1984.
75. Lastrucci, Carles L., The Scientific Approach: Basic Principles of the
Scientific Method, Cambridge, Mass.: Schenkman Publishing Co., Inc., 1967.
289
95. Pearson, Karl, The Grammar of Science, New York: Meridian Books, Inc.,
1957.
96. Phillips, Bernard S., Social Research, Strategy and Tactics, 2nd ed., New
York: The Macmillan Company, 1971.
97. Piaget, Jean, Main Trends in Interdisciplinary Research, London: George
Allen and Unwin Ltd., 1973.
98. Popper, Karl R., The Logic of Scientific Discovery, New York: Basic Books,
1959.
99. Rajaraman, V., “Fundamentals of Computers,” New Delhi: Prentice -Hall of
India Pvt. Ltd., 1985.
100. Ramchandran, P.,Training in Research Methodology in Social Sciences in
India, New Delhi: ICSSR 1971.
101. Redman, L.V., and Mory, A.V.H., The Romance of Research, 1923.
102. Roscoe, John T., Fundamental Research Statistics for the Behavioral
Sciences, New York: Holt, Rinehart and Winston, Inc., 1969.
103. Runyon, Richard P., Inferential Statistics, Philippines: Addison -Wesley
Publishing Company, Inc., 1977.
104. Sadhu, A.N., and Singh, Amarjit, Research Methodology in Social
Sciences,Bombay: Himalaya Publishing House, 1980.
105. Seboyar, G.E., Manual for Report and Thesis Writing, New York: F.S.
Crofts & Co., 1929.
106. Selltiz, Claire: Jahoda, Marie, Deutsch, Morton, and Cook, Stuart W.,
Research Methods in Social Relations, rev. ed. New York: Holt, Ri nehart and
Winston, Inc., 1959.
107. Sharma, B.A.V., et al., Research Methods in Social Sciences, New Delhi:
Sterling Publishers Pvt. Ltd., 1983.
108. Sharma, H.D., and Mukherji, S.P., Research in Economics and
Commerce, Methodology and Surveys, Varanasi: Indian Biographic Centre, 1976.
109. Siegel, S., Nonparametric Statistics for the Behavioral Sciences, New
York: McGraw-Hill Publishing Co., Inc., 1956.
110. Subramanian, N., “Introduction to Computers,” New Delhi: Tata McGraw-
Hill Publishing Co. Ltd., 1986.
111. Summers, Gene F., (Ed.), Attitude Measurement, Chicago: Rand McNally
& Co., 1970.
112. Takeuchi, K., Yanai, H. and Mukherjee, B.N., The Foundations of
Multivariate Analysis, New Delhi: Wiley Eastern Ltd., 1982.
113. Tandon, B.C., Research Methodology in Social Sciences, Allahabad:
Chaitanya Publishing House, 1979.
291