CHAPTER 1 CONCEPT OF RESEARCH 1.

0 INTRODUCTION

Research can be called as a purposive investigation. The objectives of the research could be gaining familiarity with research objectives (exploratory research), to describe the characteristics of a market or many markets as well as of consumers, to decide the fre uency with which some phenomenon (say stock!out situation for essential goods distributed as ration!card), to test hypothesis, etc. The significance " importance of the research can be understood from the fact that it provides the basis for all policies and strategies may be for any marketer or even to government. There are many types of research like descriptive, analytical, applied, basic, uantitative and ualitative, etc. 1.1 (#) (%) (&) DEFINITIONS OF RESEARCH $t is the activity which extends, corrects and verifies the knowledge. $t is the activity of finding new ways of looking at known " familiar data in order to explore new ways to change it as expected " intended. Research is the process which involves the steps like defining the problem, identifying research objectives, formulating hypothesis, collecting and interpretation of data, deriving findings, conclusions and then identifying the action plan. Research is the well planned activity which is designed and implemented to provide the data for solving important genuine and recurrent problems. Research is the activity which involves manipulation of things, concepts or symbols for the purpose of generali)ing to extend, correct or verify knowledge, whether that knowledge aids in construction of theory or in the practice of an art.a Research can also be said as a movement from the known to unknown facts $t is the systemati)ed efforts to gain new knowledge. DEFINITIONS OF METHODOLOGY AND RESEARCH METHODOLOGY

(') (()

(*) (+) 1.2

,-ethodology is a term which should not be misused for ,method. or ,techni ue.. -ethodology has got an important meaning. $t becomes first an approach towards in uiry and or research then later evolves into particular methods or techni ues. $n the applied use it is concerned with selecting specific technical tools and techni ues for collecting data and analy)ing it. $n the theoretical use, it is concerned with the philosophical fields of in uiry that can be used to conceptuali)e the problem under study. There are two different methodological stances. /iscipline research is oriented towards enriching knowledge in a scientific discipline whereas policy research denotes to another methodology that is philosophically committed and serve as a guide to social action. 0uite often methodology is used in the applied sense undermining its theoretical perspectives, though both are the two sides of a coin. 1ome of the definitions are as follows2 #

(#)

-ethodology in the applied sense refers to various methods used by the researcher right from data collection and various techni ues used for the same for interpretation and inference. -ethods and techni ues are often used synonymously in research literature. Research methodology is what must be done, how it will be done, what data will be needed, what data gathering will be employed, how sources of data will be selected and how the data will be analy)ed and conclusions reached. 3hen we talk of research methodology we not only talk of the research methods but also consider the logic behind the methods we use in the context of our research study and explain why we are using a particular method or techni ue and why we are not using others so that research results are capable of being evaluated either by the researcher himself or by others. Research methodology is ,a procedure designed to the extent to which it is planned and evaluated before conducting the in uiry and the extent to which the method for making decisions is evaluated.. The word methodology is used freely in different context. -ethodology is not merely description of methods " set of methods or techni ues. Techni ues are aids to research like are aids to research like regression and correlation. -ethodology provides arguments, perhaps relationships, which support various preferences entertained by the scientific community for certain rules of intellectual procedure, including those for forming concepts, building models, formulating hypotheses and testing theories. -ethodology is neither a study of 4good5 methods nor a study of 4methods5 used but rather a study of reasons behind principles on the basis of which various types of propositions are accepted or rejected as part of the body of ordered knowledge in general or of any discipline. $n short, we may define methodology as the science of procedure to build, verify or extend scientific knowledge. 6 thorough understanding of a scientific methodology alone will contribute an appreciable research piece. 1cientific piece of investigation will provide an argument which is as true for each individual mind as of the researcher5s own mind. Therefore, the most important step in a research design is the selection of an appropriate methodology. The significance of the use of the term 4methodology5 is that it re uires an argument to connect the choice and practice of particular methods to the way that the problem is conceived and the utility and limitations of the outcome. $t is in this sense of the term, as re uiring a critical justification for the adoption and practice of particular research methods that we claim that our concern is with 4methodology5 rather than with methods alone. .. 7nly rarely do books on research methods discuss situations in which particular methods should not be used, or situations within which the methods chosen may cause distortion or precipitate changes are the not captured by the methods themselves.

(%)

(&)

(')

-ethodologically designed research can be considered as a piece of scientific work though approximation to fool proof methodology is a continuous and never ending %

price. Thorough knowledge about the latest development in the concerned branch is absolutely necessary for designing a scientific methodology. Research is a fact finding process through systematic and indepth study through the various research process including collection compilation, presentation and interpretation of derails or data. The only way to find truth and gain knowledge is the scientific method of investigation. The researcher must be able to give clear scientific explanation and the logic behind them for a host of the following uestions2 #. %. &. '. (. *. +. :. 3hy the particular research study is undertaken8 9ow the research problem is stated8 3hat are approaches towards the in uiry8 3hat are the tools and techni ues that will be used for data collection8 3hy this method is adopted8 $s the sampling design appropriate8 9ow the hypotheses have been framed8 9ow the hypotheses will be tested8 9ow the various statistical tools and techni ues are selected. 3hat is the method of data processing8 9ow it will be calculated8 3hich techni ues are used to evaluate the accuracy of results8

The framing of a good research methodology is compared to that of an architect who designs a building, i.e., ,he has to evaluate why and on what basis he selects particular si)e, number and location of doors, windows and ventilators, uses particular materials and others and the like.. ;iterature review and interactions with experts will help one to sharpen the methodology. The external examiners who evaluate the thesis, approve one which is conceptually, methodologically and factually correct and to the best of his knowledge it is free from errors and plagiarism (copying) and sufficiently of good standard. The research uality that e uates to international excellence or national excellence in all areas or in majority of the areas of a thesis to a great extent depends on the formulation of a good and scientific research methodology. <inally, the methodology adopted should be open to pubic so that others can know how one reached the conclusions about a study. The means of method of en uiry is opened for public evaluation and criticism. 1.3 OBJECTIVES OF THE RESEARCH

The basic purpose of the research $s to identify the action plan to answer to uestions through the application of scientific procedures. The key objectives of the research could be as follows2

&

(#)

To become familiar with a certain mechanism or phenomenon or to gain new insights into it. This type of research is normally called as exploratory research design. <or example, before parliamentary elections all political parties conduct research on priorities of the voters so as to prepare election promotion plan. To describe accurately the characteristics of a particular consumer!group " segment, market situation or an industry. <or example in $ndian Tourism marketing per capita spending power of domestic tourists in Rs. &(== per annum whereas spending power of inbound and outbound tourist is Rs. (:=== p.a. 6s regards to market situation is concerned, we can describe $ndian 1oftware industry revolving around only financial solutions for banking sector. 3e can also describe $ndian home 6ppliances $ndustry as a market worth Rs. ':=== crores, growth rates +.#=> and having intense completion, but main dominant players are -?@s like ;A, 1amsung and home grown Bideocon international. To identify the fre uency with which some phenomenon, say stock!out, occurs or the causes associated with the particular phenomenon. <or example in a thick populated area, there could be fre uent stock!out of most essential goods like ;CA!cylinders, kerosene, food grains, etc. 1tate Aovt. may be wanted to know fre uency of stock out and causes like distribution bottlenecks, etc. The leading retailers of $ndia like <ood world, 1hoppers 1top, etc. conduct stock taking and or implement latest 1@- software to avoid the situations of stock out. 1imilarly many industries in $ndia and in globe, conduct research to study business! recession cycle fre uency. <or example the global recession during #DD# and #DD+ was well predicted before occurrence. $n similarly way the recession to $ndian Telecom $ndustry during %===!%==% was also predicted. @urrent predictions about some of the industries are management education, especially -E6 and -@6 is booming and will appreciate for another #= years. There is great demand for -E6s having engineering background.

(%)

(&)

(')

To formulate and test a hypothesis may be to establish the relationship between say sales and market share or sales and customer satisfaction, level, etc. <or example, Aujarat @ooperative -ilk -arketing <ederation (A@--<) was under impression that it has %=> market share for li uid milk (6mul!Ta)a) at ?agpur during %=='. $t conducted research for which hypothesis was formulated and tested. The reality was, 6mul5s market share formulated and tested. The reality was 6mul5s market share was only ##.(> at ?agpur during %=='. Fxactly reverse happened incase of @adbury5s /airy milk chocolates. $t thought that its market share during %==' is +=> whereas it was +%> across four metros and nine mini metros. 6fter #DDD, Bideocon5s market share in @TB market was on decline. $t wanted to know the cause. The research indicated that Bideocon needed rectification in brand!image. Bideocon immediately responded and appointed 1hahrukh Ghan as a brand!ambassador. MOTIVATIONS IN RESEARCH '

1.4

3hy to conduct the research8 The motivations could be as follows2 (#) Aspi !"i#$ "# %& i'& "(& )#$s&*+&$"i!, -&$&.i"s %+& "# "(& &s&! )(. <or example, till %==', the brand image of ;ux soap was 4The soap of the stars5. This positioning worked for (= years. 9owever during %==(, the brand lost the magic. The research was done by 9;;, which indicated that (a) positioning should be changed to glamour and luxury and (b) it should come out from the image product for female. 9;; repositioned ;ux like ;uxurious product that make her feel beautiful and special5 with punch line, 4Erings out stars in you5 9;; also changed celebrity from female actress to male actor 1hahrukh Ghan. Eoth things are now working to bring out ;ux from red to black and white. ;ifebuoy soap was also suffering from image, like 4Tandurasti Gi Raksha Garta hai ;ifebuoy5. The research said that repositioning and brand extension is must. 9;; repositioned ;ifebuoy from Tough 1oap and Tandurasti to Aermicidal perfume and extended to lifebuoy Aold and ;ifebuoy Clus. The obvious benefit of rising sales is the contribution of the research. /uring %=== 6/, 9;;5s all tea brands were suffering from competition due to which growth in sales became stagnant. $ndia5s tea market si)e is Rs. *=== crores, in which branded tea market occupies Rs. %(== cr and unbranded or loose tea occupies Rs. &(== cr. Research indicated that growth would be only in loose tea market. 9;; successfully launched 46#5 brand tea to snatch the customers from loose tea market, with punch line 4strong cup of tea5 and market segments focused are housewife and journalists. The philosophy used, 4due to strong cup of tea, ordinary man like housewife and or a journalist get courage to face difficult situations in the life. /20 I$"&$"i#$ "# .!)& "(& )(!,,&$1&s i$ #'& )#2i$1 "(& )#2p&"i"i#$. /uring %==(, CHA reduced the prices of the detergent Tide considerably. <or example its price was Rs. '= for (== gm. 9;; uickly responded to the change and launched the brand extension of 1urf 41urf Fxcel Elue5 for price of Rs. (= for +(= gm. To sustain the competition from band ambassador 1hekhar 1uman, 9;; opted comparably unknown faces with the statements and counter statements like Iayga!!!!!!!! ?ahi Iayyega. The well thought research by 9;; for surf Fxcel Elue is clear winner. Tide is in big problem. I$"&$"i#$ "# !pp,3 &s&! )( .# s+))&ss.+, ) &!"i'i"3. /uring %#st century, washing machine almost became house!hold item. 9owever, due to heavy traffics, habits towards consuming fast food at road side restaurants became very common. The result was the ,dust. and ,blackened cloths.. -arketers identified the priority of the consumer !!!!!!!! she wanted to get rid of 4daag5. 9;; CHA successful developed the ad!campaigns, 4/undate Raha Iayoge5, Guch Cane ke ;iye Guch /hona Cadata hai5, 4/aag 6cche 9ai5, etc. I$"&$"i#$ "# i$"&1 !"& s#)i&"!, 2! 4&"i$1 /s#)i!, 5&,.! &0 5i"( 2!i$ s" !"&1i) 2! 4&"i$1. (

/30

/40

@onsumer perception about the company and company products is very crucial to understand from marketing point of view. $f social welfare is integrated with marketing strategy then it could have cutting edge. 9istory proved that due to social welfare, the image of the company could be changed. <or example, Tisco says manufacturing steel is our second priority. 7ur firstJ priority is society welfare. 9ence Tisco say we also manufacture steel. The result !!!!!!!!!!!!! in last D= years there are no strikes and no production losses in spite of the location of factory on the border of Eihar and 7rissa where labour and labour union problems are dominant. CHA also designed 4Croject /hrusti5 for 43hisper5 and 4-r. Aold5 for all products under umbrella brand. 9;; 41hakti 6mma5, 4-obile hospital5 are winner society welfare projects, which helped to project favourable image. /60 I$"&$"i#$ "# 1&" &sp&)"!-i,i"3 5i"(i$ "(& )#+$" 3 !$% #+" #. "(& )#+$" 3. $ndia5s most successful $T trio is $nfosys, 3ipro and T@1. $ndia and $ndians are respected in globe due to the powerful brain of ?arayan -oorthy, 6)im Cremjee and -r. 1. Ramadorai, @F7, T@1. The respectability is the result of lot of hard work to deliver the value to the stakeholders. <or shareholders the respect is due to very attractive dividend and stock price. <or country, the respect is due to forex earning ability and employment generation. <or global countries the respect is due to its intelligence, price, service!delivery and out sourcing ability. $ always admire the skill of the trio and hence would like to describe, 4Real Aold of $ndian $T5. (Cer employee profit after tax for $nfosys during ='!=( was Rs. (,==,=== for T@1 Rs. ','=,=== and for 3ipro Rs. &,+=,===). Clease read article #.&( $ndia5s most 6dmired @ompanies.

1.6

IMPORTANCE OF THE RESEARCH

1ignificance of research and research leads to invention. <ollowing facts highlight the importance of the research (#) (%) (&) (') (() 1.7 Research facilitates logical or scientific thinking process which leads towards flow less strategy formulation. $t facilitates identification of 4trends5 which ultimately responsible in marketing opportunities. /ecision making becomes easier for well researched phenomenon. Research is important in solving various operational and planning problems of business and industry. $t helps understanding perception of the society about the marketer and accordingly designs the marketing strategy.

SCIENTIFIC METHOD *

The 1cientific method is theori)ing based on experimentation and is thus very close to Fmpiricism $t also needs uite a bit of logical deduction and so appears closer to the top of the graph. The essential tenets of the scientific method are (#) direct observation of phenomenaJ (%) clearly defined variables, methods, and proceduresJ (&) empirically testable hypothesesJ (') the ability to rule out rival hypothesesJ (() statistical rather than linguistic alternative justification of conclusionsJ and (*) the self!correcting process.K This is the most commonly used style of thinking in research methodology. $n this style of thinking hypotheses are proposed based on some proven theories and then they are practically tested. They re uire a considerable amount of logical deduction, but not as much as in Costulational theories. ?evertheless, the use of logical deduction is considerable. 3e shall discuss logicalK deduction further in the following section. T(& T(#+1(" P #)&ss Reasoning forms the basis of the scientific in uiry. The thought process of a scientist may be based on deduction, induction or a combination of both. ;et us understand in detail each of these processes of thinking needed for conducting and drawing conclusions from marketing research. D&%+)"i#$ /eduction is a form of inference where conclusion necessarily follows from the given premises, i.e. neither can the conclusion contradict the premises nor can it assume new premises. 6 deduction is correct if it is both true and valid. 6 deduction is true if the premises on which it is based are true, i.e., they agree with the real world. <or example, premise like Lworld is flatL is a false premise. /eduction based on such premise will also be false. /eduction is valid if it is impossible for the conclusions to be false, and if the premises it is based on is true. That means the method of drawing conclusion should be logical and valid. Truth and validity of a deduction together mean that conclusion is not logically justified (even if true) if either one or more of the premises is false or if the method of deduction is incorrect. The conclusion may still be correct due to some other premises not considered. ;et us look at an example. Cremise #2 Cremise %2 @onclusion2 Cavan is a good boy. Crathap and Cavan are friends. Cratap is a good boy.

9ere both the premises are true, but the argument that led to the conclusion is not valid and so the deduction is not valid. The conclusion may be due to some other reasons but not as a result of the given premises. ;et us look at an example where the deduction is not true. Cremise #2 Cremise %2 Reading too much dulls oneKs mind Crathap reads too much +

@onclusion2

Crathap must be dull!headed.

9ere the conclusion logically follows the premises, but premise # is a dubious statement. $f premise # is false then the conclusion is also false. ;et us look at an example in which there is a logical flaw. Cremise #2 Cremise %2 @onclusion2 /ogs do not hate water Rabid dogs hate water Rabid dogs are not dogs.

9ere premise # is correct and true in general. Eut, the rabid dogs are also dogs. $f we consider this, then the first premise is not correct. 7ne has to correct the first statement, based on the second. This leads to the invalid conclusion. Cremise $2 Cremise %2 @onclusion2 Rain is a probability if the sky is cloudy. The sky is cloudy today. Rain is a probability today.

This deduction is true and valid. 1uch deductions are made every moment by one and all and look obvious. I$%+)"i#$ $nduction is the conclusion drawn from one or more facts, bat not necessarily from facts alone. The conclusion explains the facts, but the facts just given are not sufficient to lead to the conclusion. There is a need for additional facts from the previously learnt knowledge. To illustrate, suppose Giran approaches his boss Cavan with a routine problem, but, to his shock, receives rude treatment for no mistake of his. Then Giran can based on his previous experience conclude any of the following2 6nother colleague of his might have just annoyed Cavan. -ay be, just then, his boss, 1udba, gave him a piece of her mind. -ay be a personal problem is bothering Cavan -ay be he was annoyed by a traffic jam on his way to office.

6ny of these conclusions can explain the fact that Cavan treated Giran rudely. 9owever, at the same time, the given fact cannot lead directly to any of these conclusions. These conclusions are based on some previous experience, i.e. on some other facts also. 6nd conclusions are in reality only hypotheses and need further verification to ascertain the correctness. C#2-i$!"i#$ O. I$%+)"i#$ A$% D&%+)"i#$ To explain an observed phenomena a researcher formulates some hypotheses that needs to be verified by the use of induction. The researchers then use deduction to check whether each of the hypotheses can explain the given facts completely by itself. 7nce this is done, it is :

necessary to perform empirical tests with all these hypotheses and then select a hypothesis that passes these tests. This method is described as the double movement of reflective thinking by /ewey and is adapted by @ooper and 1chindler. The process the researcher should follow is as outlined below. The researcher Fncounters a curiosity, doubt, barrier, suspicion, or obstacle, generally termed as a problem. 1truggles to state the problem2 asks uestions, contemplates the existing knowledge, gathers facts, and moves from an emotional to an intellectual confrontation with the problem Croposes hypotheses to explain the facts that are believed to be logically related to the problem /educes outcomes or conse uences of the hypotheses2 attempts to discover what happens if the results are in the opposite direction of that predicted, or if the results support the expectations. <ormulates several rival hypotheses /evises and conducts a crucial empirical test with various possible outcomes, each of which selectively excludes one or more hypotheses. /raws a conclusion, an inductive inference, based on acceptance or rejection of the hypotheses <eeds information buck into the original problem modifying it according to the strength of the evidence.

These steps are interdependent. They are also not se uentially fixed. Eased on the nature of the study some of the above steps may be eliminated or new steps may be added. S)i&$"i.i) M&"(#% A$% I"s M!8# C(! !)"& is"i)s Two major characteristics of the scientific method are validity and reliability. Balidity is a measure of the match between what the research claims to measure and what it actually measures. $n other words it measures the effectiveness of measurement in research. <or example, a people meter on a TB set is supposed to be measuring the viewership of a particular program, while in reality it only measures the number of occasions the TB was tuned to that particular channel when the program was relayed. The Television may be on but there may be no one watching it. -oreover even if there are viewers, the people!meter cannot count how many. Thus the assumption that the people!meter measures viewership validity is wrong. 9ence a research which assumes it is measuring the viewership with people meter is not valid. Balidity seems easily achievable, but as in the above example, there may be minor deviations that can easily go unnoticed. 9ence, to ensure validity one should carefully and purposefully probe into every detail of the research.

Reliability is a measure of repeatability of the research. $t is also a measure of the investigatorKs independence of the research. $n other words, if a research is reliable, then any other investigator repeating it will obtain the same results. This characteristic is also known as objectivity. 6chieving this is very difficult even in the hard sciences. 6part from validity and reliability, the scientific method has many other characteristics. 1ome of the important ones are2 ;ogical2 ;ogic is necessary in designing and following up a research process, and arriving at conclusions. 1ystematic2 The process of research is marked by thoroughness and regularity, and so it is considered to be systematic. Fmpiricism2 Research is done through observations that are based on direct sense experience. Thus it is empirical in nature.

Research should be carried out in a scientific manner to reduce the uncertainty in a situation and to ensure accuracy of the results the research yields. This needs the use of scientific method.

#=

Box 1.7 :Steps that make up the scientific method Observation: A good scientist is observant and notices thing in the world around him / her. (S)he sees, hears, or in some other way notices what is gaining on in the world, becomes curious about whats happenings and raises a questions about it. Hypothesis: This is a tentative answer to the question an e!planation "or what was observed. The scientist tries to e!plain what caused what was observed (huy#po$under, beneath% thesis $ an arranging) &ypotheses are possible causes. A generali'ation based on inductive reasoning is to a hypothesis. A hypothesis not an observation, rather, a tentative e!planation "or the observation. &ypotheses re"lect past e!perience with similar questions ((educated propositions) about cause) *ultiple hypotheses should be proposed whenever possible. +ne should thin, o" alternative causes that could the observations (the correct one may not even be one that was though o") &ypotheses should be testable by e!perimentation and deductive reasoning. &ypotheses can be proven wrong / incorrect, but can never be proven or con"irmed with absolute certainty. Someone in the "uture with more ,nowledge may "ind a case where the hypothesis is not true.

Prediction:

-e!t, the e!perimenter uses deductive reasoning to test the hypothesis. .nductive reasoning goes "rom a set o" speci"ic observations to general conclusions . observed cells in !, y and organisms, there"ore all animals have cells. /eductive reasoning "lows "rom general to speci"ic. 0rom general premises, a scientist would e!trapolate to results i" all organisms have cells and humans are organisms, then humans should have cells. This about a speci"ic case based on the general premises. 1enerally in the scienti"ic method, i" a particular hypothesis / premise are true, then one should e!pect (prediction) a result. This involves the use o" i" then logic.

Testin : Then, the scientist per"orms the e!periment to see i" the predicted results are obtained. ." the e!pected results are, that supports the hypothesis.

##

S)i&$"i.i) M&"(#%9I$ M! 4&"i$1 As C#2p! & D#" P(3si)!, S)i&$)&s 1cientific method, as the name suggest, is more applicable to the sciences than to the arts. $ -arketing Research, the decision!maker applies the methods of science to the art of marketing. Thus the reliability and validity of the method are lower when it applied to marketing research. The following are the major differences between the physical sciences and marketing research that affect the reliability and validity of research process. R&s&! )( )#$%i"i#$s $n physical sciences, an experiment is conducted under a controlled environment. <or example, in a chemical experiment, temperature, pressure, etc. are controlled to the re uired extent. $n marketing research, it is almost impossible to achieve such perfect control of all the variables. Mnlike physical sciences, marketing research does not involve inanimate, controllable factors, but involves people their behaviour, their perceptions and their attitudes, which change with time, place, presence of others at that instance, etc. These factors, being complex, adversely affect the reliability of research in marketing. 2ery low 3ow 4robable &igh 2ery high

Fi1+ & 1.: ; Fi'& P#i$" s)!,& "# 2&!s+ & "(& ,i4&,i(##% #. p+ )(!s& M&!s+ i$1 I$s" +2&$"s -easuring instruments in physical sciences provide very high accuracyJ <or instance, physicists can measure up to a #=!#( of a meter, which is a millionth of a billionth of a meter. 9owever, in marketing it is difficult to arrive at such accuracy. <or example many uestionnaires use a five!point scale to measure the likelihood of purchase. The scale is shown in <igure #.+. 1uch a scale gives only a crude measure2 moreover the meaning of the words in the given scale may mean different things to different people. This affects both the validity and reliability of the research. P& s#$!, I$"& &s"s $deally, the personal interests of the researcher should not affect research results. Eut it happens both in physical and social sciences. The extent to which the results affect the researcher is more in marketing compared to the physical sciences. $n physical sciences, research results affect only the fame of the researcher whereas in marketing research, they affect their work and thus their life in marketing research, it often happens that strong willed marketing managers may need research to support their decision, while researchers themselves may be anxious to see their organi)ation, and thus their careers, prosper. This #%

forces the researchers to push their research so that it is acceptable to their clients. $f the researcher is associated in the decision!making process as well, the personal interests of the researcher in the result increase further. I$.,+&$)& O. M&!s+ &2&$" 1ometimes the process of measurement may itself affect the result, i.e. the process researchers undertake for making the measurement may result in a change in the outcome. $n science, the affect of measurement on the result is not very pronounced except in fields like uantum mechanics. Eut in marketing research the influence of measurement on the result is very appreciable. <or example, when a family has a Lpeople meterL on its television set, it may modify their viewing habits because they know that all their viewing is recorded. 1imilarly, people participating in focus groups know that they are being observed and so they may come up with socially acceptable answers. Re! uestioning a group of respondents may also affect the results. $f a group of respondents are uestioned a second or third time, they may give different answers from what they would have given if they were uestioned for the first time. <or example, if a company has ui))ed a group of respondents about its brand before an advertisement campaign, then after this the group will start noticing the advertisements with more interest than it would have done without us ui))ing. This would change their responses in the uestioning from what the responses would have been had they not been uestioned before. The person who administers uestionnaires or conducts an interview can also influence the result by his communication skills and his knowledge of the project. 3hen the respondents approach him for clarifications the knowledge or lack of it will affect his reply. This, in turn will affect the result. 1ome times the mere presence of an interviewer may affect the result. This is more clearly explained in the chapter L$nstruments of Respondent @ommunicationL. Respondents themselves often change opinions with time. These factors affect the validity and reliability of marketing research. Ti2& P &ss+ & $n managerial decisions, timeliness becomes more important than the aptness of the decisions, per se. $n any given situation, there is no perfectly right decision, but only the most appropriate decision that can be taken in the available time. This exerts time pressure on marketing research and may reduce validity and reliability of the research results. S(# "9T& 2 G#!,s -arketing research generally aims at reaching short!term goals, i.e. helping in solving an immediate managerial problem -anagement does not aim at preserving and propagating the ac uired knowledge. $n comparison, science aims at the accumulation of a knowledge pool and uses this knowledge pool to arrive at some general theories. 7nce established, these theories remove the need for reinventing the wheel and allow the researchers to concentrate #&

on more advanced areas of research. Eut in marketing research, the knowledge once learnt is neither preserved, nor propagated. Fven if the results are preserved by a company or a marketing research firm, the results are not shared with other firms due to competition. This internal data may not be sufficient to give rise to stand!alone theories. -oreover, the firm would not spare resources to go into such KtheoreticalK research. This results in unnecessary repetition of the research. Eut increasingly, researchers are recording what was known in a project and are using it as a base the future projects. They are using more efficient methods like knowledge management for organi)ing and reusing internal knowledge. This is leading to a gradual understanding of the theoretical behavior of many issues, like consumer behavior, with respect to particular products. Di..i)+,"3 I$ E<p& i2&$"!"i#$ $n the physical sciences, cause and effect relationships can be easily identified with the help of experimentation. Eut experimentation with complete control of all the factors is impossible in marketing research. <or example, to test the effect of a new design on the sale of umbrellas, it is not possible to hold factors like weather constant. 1imilarly, when one is testing the effect of a new design on the sale of jeans, one cannot control factors like changing fashions. Thus experimentation, to its fullest extent is not possible in marketing research. T& 2i$#,#13 I$ T(& S)i&$"i.i) M&"(#% 1cientific method, since it is founded in science, derives its terminology from science. The basic terminology re uired for understanding the scientific method is given below. F!)"s A$% O-s& '!"i#$s <acts are phenomena that we believe are true. These facts do not change with the person who reports them. 7riginal documents and fact!gathering agencies are important sources of facts in marketing research. 7bservation is the process by which we recogni)e or note facts. These are experiential in nature (they are the expressions of our perception of reality) and tend to change from person to person. <or example, during a sales promotion, it may be a fact that the sales volume has not changed, but the sales staff at an outlet may give higher sales estimates based on their perception. The perception of increased sales may be due to the increased work pressure on the reduced staff. 7r it may be due to an actual increase in the number of non!buying customers visiting the outlet. Thus observations are the perceptions of the individuals based on Ktheir experience of reality,K and hence may vary from the facts. V! i!-,&s A$% D&.i$i"i#$s 6 variable is a physical or non!physical uantity that can take anyone of a predefined set of values, numerical or otherwise. $t can be defined as a formal representation of a property of entities. 6n entity is something that exists as or perceived as separate object. <or a table, #'

chair, a human etc. are entities. Fvery entity has a multitude of properties. <or example, a table has legs, wood type, feel, height, width, length, etc. 1imilarly we can consider two properties of human beings Kblood group and Kheight.K $t is usually represented by a symbol. The variables can be classified either based on their measurability or on their relationship with each other. 7n the basis of measurability the variables are of two types. @ontinuous2 The variable that takes an infinite number of continuous values is called as continuous variable. <or example, if satisfaction is represented as a variable, it can actually take any value from )ero to infinity. -athematically a continuous variable is such that, if we take any two values of a continuous variable we can find at least one more value between these values. /iscrete2 This type of variable takes only a fixed number of values. <or example, the variable 4occurrence of saleK can take any of the two valuesK # for sale and K=K for no sale, and so can be called a dichotomous variable. 1imilarly, Kdegree of likingK is referred to as a polytomy because it takes multiple values. $t can take the value K!#K for dislike, K=K for neither like nor dislike, and K#K forK like.K /ependence2 The researcher tries to establish a relationship between two variables in his research. <or sample, when he is conducting an experiment, the researcher manipulates a variable and measures the effects on some other variable. The variable manipulated is called the independent variable ($B), and the variable measured is called the dependent variable (/B). To illustrate further, suppose a researcher is trying to find the relationship between the length of an advertisement and the recall. The recall percentage is the dependent variable and the length of the advertisement is the independent variable. Table #.+ lists some terms that are used as synonyms for the dependent and independent variables. The above two types of variables are different from each other in terms of their relationship. The other types of variables, based on their relationship with other variables are the following. M#%& !"i$1 '! i!-,& /MY)2 The moderating variable is the second independent variable included in the study since it is believed to have a significant effect on the relationship between the main independent and dependent variables. <or instance, let us state a hypothesis ! the introduction of a dating allowance ($B) will lead to higher productivity (/B), especially among younger (age is -N) workers. 9ere the younger workers have a moderating effect on the original relationship. E<" !$&#+s '! i!-,&s2 These are variables outside the immediate relationship between independent variables and /ependent variable. There are many extraneous variables that have some impact on the original relationship between the $B and /B, but the effects are either not significant or so random that they are not measurable. I$"& '&$i$1 '! i!-,&2 1ometimes one finds that the $B!/B relationship stated is not direct and that the independent variable actually affects some other variable (the intervening variable or $$B), which in turn affects the dependent variable. $n the hypothesis stated above, #(

we can see that the dating allowance ($B) does not directly affect the productivity ($$ B), but affects the satisfaction in the personal life ($$B), which in turn affects the productivity (/B). D&.i$i"i#$2 /efinitions are of two major types, constitutive and operational definitions. $n constitutive definitions concepts are defined with the help of other concepts and constructs. $n other words, they are theoretical definitions. 7perational definitions are those which define concepts in terms of the process of measurement or manipulation. /efinitions are re uired in research to provide an understanding and measure of concepts. T!-,& 1.7 ; S3$#$32s .# %&p&$%&$" !$% i$%&p&$%&$" '! i!-,&s I$%&p&$%&$" '! i!-,&s @ause 1timulus Credicted from OOO. 6ntecedent -anipulated C#$)&p"s !$% )#$s" +)"s @oncepts are abstract ideas generali)ed from particular facts. They are characteristics associated with certain events, objects, conditions, situations and the like. They are used to classify, explain and communicate a particular set of observations. @oncepts are developed out of personal or group experiences over time. The concepts developed are shared between the users and thus they form the basis for the development of new concepts. @oncepts are also borrowed across fields. <or example, the concept of distance is borrowed from physical is used in attitude measurement to refer to the degree of difference between the attitudes of two people. <urther, we keep adding new meanings to the existing concepts, that is we broaden them as we ac uire more knowledge about it. Eut people teed to differ in the meanings they attribute to a concept, and this may cause problems in communication. <or example, concepts like personality, leadership, motivation, social class, etc. have a variety of meaning and so people may to perfectly understand each other when they use these words. @onstructs are highly abstract concepts. These are not directly tied with reality but are derived on the basis of other concepts. These are normally ideas or images specifically invented for a specific research or theory building purpose. The difference between concepts ad constructs can be best explained through an example. 6 maga)ine wants to check the uality of the news reports it receives on various parameters. The job has been given to Bivek, the uality consultant. Bivek finds that various attributes like news coverage, grimmer, lucidity are important. These concepts are ualitatively measurable. ?ow he finds that these concepts can be classified under some related groups. These groups can be labeled and they represent some idea or image to describe the ualitative re uirements of a news report. P #-,&2s #* D&p&$%&$" V! i!-,&s Fffect Response Credicted to OO.. @onse uence -easured outcome

There are two major types of problems in marketing research managerial problems and research problems. -anagerial problems are defined as uestions raised in a business setting. Fvery managerial problem may not re uire research. 3here there is a need for research, one needs to define research problem. Research problems are the restatements of managerial problems so that the researcher understands the problem the decision!maker is facing. The objective of marketing research thus becomes solving the managerial problems by finding a solution to the respective research problems. /rink$t, a soft drink company with a strong trendy image, now intends to target the older generation in order to expand its market. They have several alternatives before them. They can either shift the brand image and project it as a soft drink for all ages or they can introduce promotional campaigns with the message that the older generation can project themselves younger by consuming /rink$t. 7r they can introduce a new product for the older generation, with different packaging and a different brand name. This is the managerial problem. ?ow the research problem can be stated as, L3ill the older generation like to project themselves younger8 9ow will the younger generation react to each of these alternatives8 3ill too many brands create confusion8 L The research problems are uestions about the interaction between two or three variables or concepts. To further analy)e these problems, a hypothesis is prepared.

H3p#"(&sis 9ypotheses are conjectural statements of the relationship between two or more variables that carry clear implications for testing the stated relations. They further classify research problems into statements which can be tested. These can be considered as probable answers to the research problem. $n the above example, the hypothesis statements can be as follows2 The older generation feels that projecting themselves as young means accepting that they are old. The older generation feels that they should be accepted as they are. The younger generation will lose interest in a drink that is meant for all ages. Ceople who consider themselves as neither young nor old may get confused with two brands.

T3p&s #. H3p#"(&s&s /escriptive 9ypotheses2 These are propositions that describe the state of a variable. <or instance, one can hypothesi)e that the market share of -aruti is more than (=>. #+

Relational 9ypotheses2 Relational hypotheses describe a relationship between two variables with respect to each other. The dating allowance example used earlier is of this type. 1imilarly, an increase in market share (/B) due to improved features ($B) is an example of a relational hypothesis. There are two types of relational hypotheses, @orrelational and Fxplanatory hypotheses. @orrelational Relationships state that two variables occur together without implying that one is the cause for the other. $n such cases, the two variables occur together, but we do not know of any other variables that might cause both of them. 6lso, we have not developed enough evidence to make a strong claim. <or example, we have a hypothesis stating that monsoon and demand for coffee are directly proportional to each other. These two occur together, but we do not have any evidence to prove that one of them causes the other. The explanatory or causal hypothesis have two variables interrelated to each other such that one implies the other. L!5s 7nce a hypothesis is verified by numerous researchers, in different situations, the relationship between the variables may be considered a law. 6 law can be defined as a well!verified statement of relationship about an invariable association among variables. $n business, we do not have many well!established laws, as the relationships are not fully invariable. 3e only have tentative laws that are the only to some extent. This is because of the presence of a number of $$Bs and -Bs in real situations. T(&# i&s A$% M#%&,s 6 theory is a set of statements that explains or predicts a phenomenon of interest. These statements may be facts, concepts, constructs, hypothesis or laws. Theories are always grounded in reality. 6n imaginative statement can at the most be called a hypothesis, but not a theory. These theoretical statements guide marketing researchers in conducting future research, and they also guide managers in decision!making. Theories narrow down the range of facts that researchers need to study. They also suggest methods for tackling a problem that are likely to yield the most accurate results. 1uch theories may also provide or suggest a system in which researchers can fit in the data, classify and analy)e it. Theories are also useful for predicting facts that need to be found. $f a system is represented in terms of symbols or physical analysis with the purpose of simplifying the understanding, testing and analysis of it, it is known as a model. They represent phenomena through the use of analogy and can thus explain theory better. There are three major functions of a model description, explication and simulation. /escriptive models seek to describe the behaviour of elements in a system where the theory is inade uate or non!existent. Fxplicative models are used to extend the application of well! developed theories or improve our understanding of their key concepts. 1imulation models go beyond the goal of clarifying the structural relations of concepts and attempt to reveal the process relations among them. 1.= STEPS IN THE RESEARCH PROCESS #:

The steps in marketing research process are given in <igure#.:#. These steps are interdependent and simultaneous, though they are treated here as if they were se uential. <or example, data collection methods are often dependent on the choice of sample and the analytical approach to be used. 6t the same time, the analytical approach itself is dependent on the data collection method. Thus the interdependency of these two steps re uires them to be simultaneous. Thus the steps given here are only indicative of the possible major steps that can be taken in a marketing research project and do not represent the exact se uence of the steps that are taken. This se uence varies for one problem to another. S"&p #2 D&.i$i$1 T(& R&s&! )( P #-,&2 !$% i%&$"i.3i$1 &s&! )( #-8&)"i'&s Croblem identification in a research project is like choosing the destination a journey. 6 research project without identifying the right problem is as meaningless as a journey without a destination. 6 problem as presented to the researcher is only a tentative problem, and it is just a statement of the problem as perceived by the decision!maker. $t may not be the real problem in most cases, and it is the duty of the researcher to identify it correctly. Eut the researcher cannot identify the problem on his own because he does not have all the information that the decision!maker has, and hence he needs the active participation of the decision!maker in problem identification. Eut in most cases the decision!makers may not be willing to give the complete details of the problem to researchers, either because they do not see the need to do so or because they would like to maintain secrecy. <urther, some of the decision!makers perceive problem identification as the duty of the researchers and do not see any role for themselves in it. The reluctance of the management to discuss the problem and the lack of initiative on the part of a researcher often leads to either incorrect or partial problem identification. 7bviously, if the research process is continued with such a defective problem identification and statement the results will not be of any use. This financial human and temporal resources used in that research would have been wasted. To avoid such wastage, the researcher should identify the problem in the first instance itself. Thus one can consider that in marketing research, problem identification is one of the most important steps. 3e can even say that right problem identification is e ual to research half done. To identify the right problem and understand all its dimensions, the researcher should ideally know the following. The complete situation faced by the decision! maker The alternatives he can choose from The expected outcome of the alternatives The objectives of the decision!maker.

Eut as we have already seen, the researcher is not provided with all this information. 1o, it is up to the researcher to use appropriate techni ues to collected the needed information <or #D

example, a researcher can analy)e the problem statement given by the client word by word. This will bring out the real objectives of the client.

/e"ining the 5esearch 4roblem and identi"ying research ob6ectives

7ost/2alue Analysis o" the .n"ormation, 0ormulation and testing o" hypothesis Fi1. 1>71; T(& s"&ps i$'#,'&% i$ "(& R&s&! )( P #)&ss Selection o" the /ata 7ollection *ethod

Selection o" the Sample

Selection o" the *ethod o" Analysis

8stimate the 5esources needed

4repare the 5esearch 4roposal

/ata 7ollection

/ata Analysis %=

5eporting

r%
<

r#

C P Cerifery of circle as market data r#, r% P radius of circle as research objectives < P <ocus of circle as research problem The researcher needs to understand not only the problem but also the objectives of the management. Then alone the researchers can align the research objectives with that of the managerial objectives. Towards this end the most important re uirement in research process is the communication between the researcher ad the decision!maker. Eetter the communication between them, closer the problem statement will be to the actual problem. The problem will be further clarified, if the researcher develops a situation model. 6 situation model is a description of variables and their relationships to the outcomes. Sp&)i.i)!"i#$ #. i$.# 2!"i#$ &*+i &2&$"s $nformation re uirements can be derived once the research objectives are clearly established. Fven at this stage, the management and the researcher need a good amount of communication between them so as to avoid collecting irritant data or missing out the re uirement data. %#

Researcher should consider the data availability, data collection techni ue, and ability of sample to yield the re uired data and the techni ues of analysis to be used to decide on the re uirement of the data. 6 common temptation is to collect ,all. the possible data. $n aiming to collect all the possible data there are chances of missing some important data elements and including some irrelevant data elements. -oreover, the data elements collected with such an unclear objective as all the possible data will not be focused and so will not serve any purpose. The researchers can ensure that the data collected is indeed relevant by asking uestions concerning the possible findings, due to each of the data elements. They should then trace the implications of each of these findings on the decision. $f the findings do to affect the decision then the data element that lead to the finding, should be dropped. S"&p 2; C#s" ? V!,+& A$!,3sis #. "(& i$.# 2!"i#$ .# 2+,!"i#$ !$% "&s"i$1 #. (3p#"(&sis The major cost constraints in marketing research are time and money. These costs are justified only by the value of information which is the result of the research. The decision! maker depends on research for some additional information that can reduce the uncertainty about the situation. Eut the additional data does not have any value if it is not supplied in time. 1o the time factor is important in research. Thus time can be considered major resources for marketing research. 6t this stage of research, only a rough estimate can be made of financial d time costs. ;arger the sample to be taken, the larger the costs in observational of experimental studies, because per day costs are going to remain the same. 9owever, if one tries to reduce the time period in uestionnaires and interviews, the costs will increase, often exponentially. 6lso, the uality of the output may come down exponentially again. There are two methods to estimate the value of research. $ntuitive method, the first of the two, relies entirely on the judgment of the decision!maker. The second approach, known as the expected value approach, uses Eayesian statistics to ualify the judgment probabilities. $n both these methods, certain considerations should be taken into account to estimate the probable value of the research. The following are the essential considerations for every problem. 6part from these there may be some considerations specific to the problem. The possible outcomes and their pay off2 3hen a problem is being considered, if the payoffs in each of the possible outcomes are not very different from each other, then it does not matter which one is chosen. $n such cases, the value of the research is very low. The higher the difference in pay off of various outcomes, the more valuable the information becomes. The degree of uncertainty in the situation2 Research is done basically to reduce the uncertainty connected with a decision. $f the decision!maker does not perceive much uncertainty in a situation, then the research does not have much importance. Thus the research is more valuable in a situation where the uncertainty is greater. %%

The ability of research to reduce the uncertainty2 <rom the above discussion it is cleared that higher the ability of research to reduce the uncertainty, the greater is its value. The risk presences of the decision!maker2 /ifferent firms and different individuals have different risk preferences. $f an organi)ational culture is such that it prefers more risk, it will not value the research greatly. $ndividuals who are more risk averse value the research more than individuals who are risk takers. Thereafter appropriate hypothesis may be framed and tested.

S!"#S O$ POO% P%OB&'( )'$!#T!O# 4oor problem "ormulation will come to light during the mar,eting research e!ercise. Three main signals indicate poor per"ormance in this tas,. 'xtensive !teration +ne early mani"estation o" inadequate problem de"inition in a research e!ercise is e!tensive iteration and reviewing o" study proposals. 9ith a growing uncertainty and lac, o" con"idence in progress, mangers retreat to the e!ploratory research and to repetitive des, research. They begin to reread the study proposal in an attempt to regain the conviction that their particular line o" inquiry is correct. )ifficu*ties in draftin research !nstruments Another induction that all is not clear problem statement terms is when considerable di""iculties are e!perienced in dra"ting suitable questionnaires or other research instruments. The right questions are not as,ed and this is not noticed until a"ter the survey. 3ac, o" purpose and "ocus causes time:wasting and "rustration. )ata (inin in +na*ysis 9hen it comes to that analytical stage o" a study that was originated in poor de"inition, it is "ound that substantial (data mining) is called "or an endeavour to discover (interesting) relationships. The absence o" clarity and purpose in collecting data means that its content, depth, and "ormat might to be correct, or it might. 8!haustive searching "or meaning may uncover use"ul relationships in the data, or it might not;.. <ltimately the de"inition o" the problem is a ,ey determinant o" the e""ectiveness o" technique application, the quality o" the proposed solution and the mar,eting management decision in turn. ::::::::::::::::::::::::::::::::::::::::::::::::: Source www.mcb.co.u,

S"&p 3; S&,&)"i#$ #. "(& %!"! )#,,&)"i#$ 2&"(#% 6 researcher can use two types of data2 primary data, data exclusively collected for the current problemJ secondary data, data collected for some other purpose and which is useful in the present research. $n an exploratory survey, secondary data is used more regularly because of its cost and time advantages, whereas in conclusive research the usage depends upon the case. %&

The data can also be classified into survey data and experimental data depending on the method of collection used to collect it. These methods, i.e. the survey method and the experimental method, can be sources of primary and secondary data.

T!-,& 1.= ; M!8# D!"! C#,,&)"i#$ 2&"(#%s I. S&)#$%! 3 R&s&! )( Mtili)ation of data that were developed for some purposes other than for solving the problem at hand. a) $nternal secondary data /ata generated within the organi)ation itself, such as sales person call reports, sales invoices and accounting records b) Fxternal secondary data /ata generated by sources outside the organi)ation, such as government reports, trade association data and data collected by syndicated services II. S+ '&3 R&s&! )( 1ystematic collection of information directly respondents a) Telephone interviews @ollection of information from respondents via telephone b) -ail interviews @ollection of information from respondents via mail or similar techni ues c) Cersonal interviews @ollection of information in a face!to!face situation. Cersonal interviews in the respondent5s home or 9ome interviews office $ntercept interviews Cersonal interviews in a central location, generally a shopping mall d) @omputer interviews Respondents enter data directly into a computer in response to uestions presented on the monitor Crojective techni ues and depth /esigned to gather information that respondents interview are either unable or unwilling to provide in response to direct uestioning. Fxperimental Research Researcher manipulation of independent variables in such a way that its effect on one or more other variable can be measured. ;aboratory experiments -anipulation of the independent variable (s) in an artificial situation. <ield experiments -anipulation of the independent variable(s) in a natural situation. %'

S"&p 4; S&,&)"i#$ #. "(& s!2p,& -arketing Research, as discussed in the chapter ,$ntroduction to marketing research. has come into existence basically because of the vast si)e of the market. /ue to this si)e, it has become impossible to collect information from the entire population of a target market. 1ampling is used to overcome this problem. The other reasons for the use of sampling are given in the chapter ,1ampling.. There are various decisions a manager should take to arrive at a sampling plan. They are as follows2 Copulation Q determines what forms the population that provides the information. 1ampling Mnit Q determines the individual sampling unit. (Cersons, households, companies and city blocks, etc.). This is treated as an individual unit in the process of sampling. 1ampling method Q determines the method of sampling. Crobability Q sampling units are selected at random and there is a known probability of each unit being selected. ?on probability Q sampling units are selected on the basis of convenience or judgment, or by some other means, and so one cannot allocate to a unit a particular probability of being selected. 1ample si)e Q determines the si)e of the sample to be used. This is based o the time, cost and necessary precision. S"&p 6; S&,&)"i#$ #. "(& M&"(#% #. A$!,3sis 6 given method of analysis re uires a particular data element in a particular format. 1ince each analytical method re uest different data elements, in different forms, it is necessary to determine the analytical method before venturing into data collection. 6gain, the method of analysis depends on the nature of the sampling process and the data collection method. 1o, decisions about the data collection method and the method of analysis should be taken simultaneously. 7nce the analytical methods have been selected and proposal approved, the researcher should design the response instruments and generate dummy data. /ummy data is the hypothetical data generated imaginatively by the researcher to check whether the analysis techni ues are working as they should. This imaginative data has all the characteristics of original data. The dummy data should then be fed into the analysis tables and checked for completeness. The analysis will also expose any redundant data!elements in the original plan, and it will also reveal any missing essential data!elements. S"&p @; Es"i2!"& "(& R&s#+ )&s N&&%&% The resources needed for research are time, finance and personnel. Time and financial re uirements are inversely dependent on each other i.e. if one likes to reduce financial %(

expenses then the research may take a longer time period, and if one tries to conduct the same research in lesser time period the financial expenses may increase. Researchers need to strike a balance between the use of these two resources. <inancial costs include direct and indirect costs like manpower, materials, overheads, etc. -any organi)ations have a rule of thumb for estimating the cost. <or example, they can have Rs. R as fixed cost and Rs. N as variable cost for each interview. 1uch formulae help in faster estimation of the resources needed.
Box 1.,- : '&'('#TS O$ TH' %'S'+%.H P%OPOS+& 8!ecutive Summary = a brie" statement o" the ma6or points "rom each o" the other sections. The ob6ective is to allow an e!ecutive to develop a basic understanding the proposal without reading the entire proposal. >ac,ground = A statement o" the management problem and the "actors that in"luence it. +b6ectives = a description o" the types o" data the research pro6ect will generate and how these data are relevant to the management problem. A statement o" the value o" the in"ormation should generally be included in this section. 5esearch Approach = a non:technical description o" the data:collection method, measurement instrument, sample ad analytical techniques. Time and cost requirements = an e!planation o" the time and costs required by the planned methodology accompanied by a 485T chart. Technical Appendi!es = any statistical or detailed in"ormation in which only one or a "ew o" the potential readers may be interested.

S"&p :; P &p! & "(& R&s&! )( P #p#s!, 6 written research proposal puts down on paper the management problem, the research objectives, the research methodology and the resource re uirements. This will help the researcher and the decision!maker to be in perfect agreement with each other to the extent that they derive the meaning from the same words. This will ensure that the research is on the right track. Eox #.:% gives a complete description of the various elements of the research proposal. S"&p 7 ; D!"! C#,,&)"i#$ /ata collection re uires trained people for ensuring the validity of the research. $f a firm5s re uirement does not warrant a permanent team, it can outsource personnel from data collector suppliers. 3hen the firm hires data collectors, it should take care that these hires are well trained. -oreover, data collectors, whether external or internal should be given a complete picture of the research before they are assigned a task. This will reduce errors as the data collector5s interpretation will be in tune with that of the researcher. Research training %*

and evaluation of field workers can also help standardi)e data collection methods and reduce errors.

S"&p = ; D!"! A$!,3sis. Eefore the data is collected is analy)ed, it needs to be edited, coded and tabulated. 7nce it is tabulated the data is analy)ed using statistical analysis. The results obtained from analysis are interpreted to some extent by the researcher. The rest of the interpretation is done by the decision!maker himself, as he understands the problem situation more clearly than the researcher. The reliability of the analysis is estimated by error estimation methods. 7nce the analysis is done and the interpretation is made, the researcher needs to report the research to the decision!maker. S"&p 10; R&p# "i$1 Reporting is the culmination of a research effort. 1ince it involves communication, one should take care of the factors affecting communication. This means that the report should contain both technical detail and managerial implications. $t should consist of an executive summary that mentions the managerial implications. This should be then supplemented by the technical details, so that the decision!maker can refer to them as and when needed. The report should also cover the methodology used in the research. 9owever, one should remember that a written report might not really invite action unless the management is very interested in it. 1uch an interest can be generated only if the manager is involved in the research from the beginning of the project. 6lso, many managers do not respond to the written word. 1ome managers may respond, but may misunderstand the written material. 9ence the written report must be supplemented with an oral report. This oral report, depending upon the situation, can range from a briefing to a full!fledged audio!visual presentation to an executive body. 1.10 REPORT ARITING

1teps in writing report (a) ;ogical analysis of the research objectives (b) Creparation of the final outline on findings (c) Creparation of rough draft (d) Creparation of final " corrected draft 1.101 L!3#+" #. R&s&! )( R&p# " (a) Creliminary pages Title 6cknowledgement Creface " <oreword @ontent %+

;ist of tables, figures 6bbreviations used (b) -ain Text $ntroduction 1tatement of recommendations 1ummary " 1ynopsis (c) Fnd -atter $ndex Eibliography 1ample uestionnaire ;ist of sample @ompany Croduct " 1ervice findings, conclusions and

#.#=% T3p& #. &p# " (a) Technical Report (b) Copular Report E2p(!sis #$ (i) -ethods and research (ii) 6ssumptions (iii) 6nalysis of findings 7bjectives, findings and recommendations (mathematical part is avoided)

1.103 Ess&$"i!, *+!,i"i&s #. &s&! )( &p# " (#) (%) (&) (') (() (*) (+) (:) (D) (#=) $t should have ade uate length to cover the research subject. $t should maintain interest of the reader. <or this big paras as part of discussions to be avoided. 6bbreviations to be avoided. Readers are interested in uick knowledge. 9ence in the beginning of report, the findings should be highlighted in executive summary ;ayout must be as per research objectives. ?o grammatical mistakes 6ll figures must be named, analysis must be in structured manner $t must show originality. $mplications of the findings to be discussed Report must be attractive i.e. clean, neat in appearance

%:

6bove format is used mostly by western corporates. The type of the format $ndian @orporate use is given in last chapter. Research Croblem Q To study market penetration of 1urf /etergent Cowder in H around Cune and to recommend marketing strategy. Research 7bjectives To find out2 (a) (b) (c) (d) (e) 1tep % 2 (a) (b) 3hich is the most commonly used detergent in the market8 3hat influences people to buy a particular brand8 3hat is the penetration level of 1urf in the market8 To identify customer needs. To design marketing strategy /eveloping research plan Research /esign Q /escriptive 1ources of secondary data Q <rom $ndian retiailers association, name and addresses of grocers, supermarkets were collected, from whom name and addresses of detergent users were collected. 1ources of primary data Q 9ousehold samples Research approach Q <ocus group interviews and individual sample interview. Research instruments Q 1tructured uestionnaire

(c) (d)

(e)

1ample Clan

Mniverse Q Residents of Cune <rame Q /etergent users 1ample si)e Q (% " '= ((% list of detergent users from whom '= 1urf users were picked up) -ethod Q 1ystematic sampling Mnit Q 9ousehold

(f)

@ontact method Q T$ " C$ 0uestionnaire for households samples. /ear 1ir " -adam,

The students of -anagement studies, Cune are conducting this survey, as a part of their project in the field of -aret Resarch. The purpose of this activity is to measure the penetration of 1urf in Cune. %D

#. %. &.

'.

(.

*. +.

:.

D.

/o you wash your clothes at home8 Nes ?o /o you use 1urf8 Nes ?o $f 1urf, which sub!brand do you use8 1urf Fxcel,1urf Fxcel Elue 1urf Mltra 1urf 1uper Fxcel 1urf Fxcel -atic 3hat influences your decision while buying 1urf8 (Tick as many as applicable) 3hiteness ;ather Fasy on hands Fasy on fabric 1tain removal 6ny other (please specify) have you seen any promotional campaign of 1urf8 Nes ?o $f yes, which one do you like the most8 ;alitaji 1urf Fxcel hai na /ho daala /hoondhthe Reh jaaoge 6ny other (please specify) SSSSSSSS /o the various schemes associated with 1urf affect your purchase8 Nes ?o 3ould you suggest any changes for 1urf in the following fields8 6vailability in different uantities 1tyle of packaging -ore schemes to be associated with the brand Cricing 6ny other (please specify) SSSS 3hy not 1urf8 Crice 0uality Cackaging <ewer schemes as compared to other brands 6ny other (Clease 1pecify) S 3hich detergent do you most frfe uently use8 (Tick as many applicable) 6riel ?irma 3heel &=

 Rin Tide 9enko 6ny other (please specify) SSSSSSS #=. 3hat influences you to buy your preferred brand8 <riends ?eighbours 6dvertisements 1elf Fxperience 6ny other (please specify) SSSSSS ##. 3hile purchasing a detergent, what uantity do you usually go for8 ;ess than # Gg #!% Gg %!& Gg &!' Gg -ore than ' Gg #%. how fre uently do you purchase detergents8 7nce a week 7nce a fortnight 7nce a month 7nce in two months #&. Nou prefer your detergent in2 1achets (#= gm, %= gm, (= gm, etc) Cackets Iars Eigger containers 6ny other (please specify) SSSSSSS #'. $f your preferred detergent is not available, you go for2 <irst @hoice SSSSSSSSSSSSSS 1econd @hoice SSSSSSSSSSS #(. do you keep a stock of detergents in your home8 Nes ?o #*. -ost preferred detergent among people you know 1urf 6riel ?irma 3heel Rin Tide 9enko 6ny other (please specify) SSSSSSS S#2&"(i$1 !-#+" 3#+ ?ame 2-r."-rs."-s. SSSSSSSSS 6ge Aroup 2 Gindly tick whichever is applicable T %( &#

 6ddress2 7ccupation2

%(!&' &(!'' '( and above

/o you own a washing machine8 Nes ?o 3ho washes the clothes in your house8 Nourself -aid 6ny other (please specify) SSSSSSS 9ow many members are there in your household8 SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSS $ncome Aroup 2 (Tick whether is applicable) T (=== (,==#!#=,=== #=,==#!#(,=== #(,==# and above Thank you 1tep &2 <orming temporary marketing organi)ation for collection of market data Croject ;eader ! # -R 7fficer (?ot re uired) $nvestigators Q # Time to complete the project Q since (% " '= samples to be interviewed and one sample might take &= minutes and &= minutes could be consumed in traveling, in one day, : samples could be interviewed. 9ence project will be over on (th day. ?o. of investigators needed is only one . 1tep ' H ( 2 Research Report. /ata analysis by using 1C11. /ata presentation and preparation of

&%

S!2p,& C#2p#si"i#$ $n all, group members as a part of our survey visited (% households. #% of them revealed that they were entirely dependent on local washermen or launderettes. Therefore these respondents were not considered for answer the uestionnaire. The remaining '= thereby formed the sample si)e of our survey.

Sex

*ale 0emale

O../P+T!O#
B@ A@ C@ @ &ousewi"e
FE@@@ E@@C:C@@@@ C@@@C:CE@@@ GCE@@@

AB C@

Sample si'e $ ?@
#

!#.O(' "%O/P
H

Student

9or,ing

Sample Si'e $ ?@ C?

C@ C@ @ A ? +"' H C? I C@ "%O/P CA E CA C? CH Sample Si'e $ ?@

CE C@ E @
FAE

AE:B?

&&BE:??

G??

Sample Si'e $ ?@

AE A@ CE C@ E @

AA CA H

Joursel" C
A B ? E H

*aid / servant AIK

('(B'%S !# HO/S'HO&) Sample Si'e $ ?@ +thers

EK C@K C?K

O0# + 0+SH!#" (+.H!#'1

A?K E@K CDK E@K

Sample Si'e $ ?@

Fi$%i$1s #. #+ S+ '&3

Jes

-o

Sample Si'e $ ?@

ABK

##K 6s a part of our survey, we visited (% houses. $t was found that #% households gave all their Sample Si'e $ EA at home. 1ince the clothes to launderettes, while '= Jes households washed their clothes -o &'

objective of our survey was to find out which detergent is popular in the households, we did not take into consideration the #% who depended entirely on launderettes.

EAK

?IK

7f the '= people interviewed, we found that %# households used 1urf, while #D of them washed their clothes Sur" with other detergents. This is a clear indicator of the popularity and the presentation of this particular brand in the consumer5s in the market. Samplemind Si'eand $ ?@
+thers

3.

S+-9- !$%s #. S+ . +s&%

A@ CE C@

C#

A A @ 7f the %# consumers using 1urf, it was found that 1urf Fxcel H Elue as a sub!brand was the @ most commonly sued, with #+ consumers it as their preference. This was followed by Sur" 8!cel Sur" <ltra stating Sur" Super Sur" 8!cel 1urf Mltra and 1urf 1uper Fxcel with % consumers each. 9owever, no users could be detected 8!cel *atic o" Sur" AC for 1urf Fxcel<sers -atric.

'.

$nfluential factors while buying 1urf EK C?K B#K

9hiteness 3ather easy on hands easy on 0abric Stain 5emoval +thers

A?K &( C@K C@K

Sample Si'e $ AC

6s indicated above, whiteness that the detergent provides, say : of the consumers, is one of the most potent influences while buying the detergent. The second most important influence is the fact that it is easy on the fabric, say ( of them. 7ther influential factors are its gentleness on hands and its good stain removing capacity (/aag dhoondte reh jaaoge). 6. C#$s+2& !5! &$&ss 5i"( &sp&)" "# "(& !%'& "isi$1 )!2p!i1$ #. S+ .. C@K
3alita6i

A?K

BAK

Sur" 8!cel hai -aa

C@Krespect to advertising A?K 7f the consumers /ho surveyed, by 1urf was cent per cent! /aala awareness with that is, all consumers using 1urf were aware of its promotional campaigns and all ahd been /hoondte 5eh 1urf ads at one point of time or the other. 7f the different tpes of ads aired by the media, the Laaoge lalitaji ad held the Any greatest retention power and linking, with + out of the %# consumers liking +ther it the most, followed by the ad for 1urf Fxcel and /hoondthe Reh Iaaoge, with a fan following ( consumers each. Sample Si'e $ AC *. Cersuasive Cowers of various 1chemes, which are launched by 1urf to promote sales, are generally not T9F major criteria when the consumer goes in for a purchase. This is also reflected by the survey in which #& out of %# of the consumers supported the fact. 7nly : were those who were affected by the schemes propagated by 1urf. +. 1uggestions provided by the consumers

The following change were suggested in the 6ny other category. :. /emands for a measuring scale so as to avoid wastage of powder. @hange in the colour of the detergent powder. <ragrance of the detergent. Reasons for not using 1urf.
E ? B A C @ 4rice Muality

&* 0ewer 4ac,aging +thers Schemes Sample Si'e $ CD

Crice of the detergent and association with fewer schemes were the two primary reasons for which consumers preferred other brands to 1urf. Crice was a factor for users of cheaper washing powders such as ?irma, Rin and 3heel. Msers of 9enko, Tide and 6riel insisted that the uality of their detergent was superior to that of 1urf.

D.

/etergents (other than 1urf) fre uently used by consumers

@ +thers Tide C C 5in 9heel # # A A ? H I

6mongst many existing brands available (excluding 1urf) in the market, the most fre uently A used ones are Rin and 3heel followed by 6riel with others (;ocal) constituting the rest of the Arielindicted above, #D were nonusers ? market. $n the chart of 1urf, while ( of them also preferred an additional detergent besides 1urf. Sample Si'e $ A? #=. -ajor influencers while making a purchase <actors ?o. <riends & ?eighbours ' 6dvertisements #& 1elf!experience #D 7thers # Total '= 3hile conducting the suryve, personal experience of using the product along with many others over a period was major influence while indulging in the purchase. 6part from this, effective advertising was a close runner!up and was largely responsible in influencing people while buying their preferred brand. ##. 0uantity usually purchased ?o. #= %= * & &+

0uantity ;ess than # kg #!% kg %!& kg &!' kg

-ore than ' kg Total

# '=

6s is predictable, due to the fact that the survey was done in an area which was a middle Q class one, the housewives usually went in for the #!% kg pack and the fre uency of purchase was once in a month which is depicted in the cahrt above.

#%.

<re uency of purchase ?o. & ## %& & '=

<re uency 7nce a week 7nce a fortnight 7nce a month 7nce in two months Total

Msually households preferred to buy their stock of detergent once in a month, as is mostly the case with all stock being ordered along with the ration that comes monthly. Eut still many households also buy it fortnightly. Cackets ((== gm, # kg, % kg) are the outright winners in this section with more than (=> consumers in this category preferring this particular style of packaging. 9owever, Iars were also preferred because of their multi!utility purpose after using the primary product. #'. 6lternative brand of detergent N#. D : #= # + ' # '= 1econd choice B !$%s Rin 3heel 1urf Tide 6riel 9enko ?irma Total N#. ## #% ( * & # % '=

<irst choice B !$%s Rin 3heel 1urf Tide 6riel 9enko ?irma Total

There were #= non!users of 1urf who preferred it as their first choice of purchase in case of non!availability of their preferred brand. Msers of 1urf voted for 6riel, Rin and wheel as their -o first choice given the same situation. #(. 1tock of detergents
Jes

&:
@ E C@ CE A@ AE

Sample Si'e $ ?@

-ore than (=> of the households did not keep a stock of detergents at home and resorted to purchase only when the need arose. #*. Creferred detergent amongst ac uaintances of consumers

The general impression that we get after conducting the survey is that 1urf rules the market because it was revealed that amongst the ac uaintances also 1urf was the most popular brand followed by 6riel and ?irma. Fi$%i$1s $n 0.?o. +, eight respondents stated that they would like more schemes to be associated with 1urf. 9owever, when they were asked that what change would they suggest in their detergent (0.?o. :), only ( of them suggested more schemes. 1urf Fxcel (#+"%#) is preferred by the consumers because of its extraordinary whiteness (:"%#) and the fact that it is easy on the fabric (("%#). 3hen it comes to housewives the verdict is almost e ual with #% saying 4Nes5 and ## saying that they do not use 1urf but when it comes to students, 1urf is the clear winner with * out of + favouring the product. 7f the '= consumers surveyed, %# 1urf and of those %#, #D were women as 1urf is more a product that homemakers use. 7f the #D non!users, #* again were women with the rest being men who had genuine knowledge about the product and who had used it at one moment of time or another. 7f the users of 1urf, all of them were more or less e ually distributed when categori)ed according to the income group with the higher income group categories preferring 1urf a little more as 1urf is costlier than most of the other brands (#&"%#). 6n interesting fact is revealed, ' suers stated that some of their clothes were washed either by themselves or by their maids, however the expensive clothes were given to launderettes. 3e also find that the trend of people who are in the different categories is almost the same with almost an e ual number in each category. &D

7ne interesting observation may be possible. $t is seen that the less than %( age group of users are more inclined to sue 1urf and as the age group increases the number of users decrease, this may be due to the new positioning that 1urf is using where it is targeting the younger generation too, through its advertisements. 1urf is popular with ac uaintances of both the users and the non!users. $n the ac uaintances of users section 6riel follows (("%#) while in the alternative category ?irma (*"#D) and 6riel ('"#D) are preferred widely.

R&)#22&$%!"i#$s .# %&si1$i$1 2! 4&"i$1 s" !"&13 7f the sub!brands, 1urf Fxcel was the most recogni)ed one, so the company ought to take some measures to make the consumer aware about other sub!brands. $f possible, pricing should be reviewed, with many consumers citing it as a negative factor. 1urf being viewed as a premium product could come up with a lower priced sub! brand for more rural market penetration to compete with 3heel, ?irma and so on. -ore schemes should be introduced to attract non!users. 6dvertising standards should be maintained, if possible improved, as advertisements have contributed immensely to the awareness level and usuage of the product.

'=

 '#

'%

CHAPTER 2 TYPES OF RESEARCH9DESIGNS 2.0 INTRODUCTION

$n the previous chapter we examined the various steps in the marketing research process. 3e also discussed the benefits of the research design. $n this chapter we will analyse two types of research designs vi). exploratory and conclusive. -arketing research projects can be classified into two major categories. Fxploratory research and conclusive research. Fxploratory research helps in discovering new relationships while conclusive research helps such situations. /ecision!makers use exploratory research for a preliminary investigation of the situation. Thus the purpose of an exploratory study is to provide new insights into a confusing issue in choosing the best from various possible courses of action. This chapter makes a detailed study of the above mentioned research techni ues. 2.1 USE OF EBPLORATORY RESEARCH

Fxploratory research emphasi)es the discovery of new ideas. Through exploration researches develop concepts more clearly, establish priorities develop operational definitions and improve the final research design. 0uite often managers face situations that are vague in nature. They may not be able to understand whether a situation present an opportunity or poses a problem for them. Fxploratory research is ideal in dealing with such situations. /ecision!makers use exploratory research for a preliminary investigation of the situation. Thus the purpose of an exploratory study is to provide new insights into a confusing issue. Msually the researcher studies the situation and identifies the main factors contributing to that particular situation. 6s the number of the factors affecting the bottomline of an organi)ation may be large, the researcher then converts these factors into specific hypotheses relative to possible actions. These hypotheses are then tested by conclusive research .<or instance, a television company may notice a change in the sales figures. Eut they may not be able to pinpoint the factors that affect the sales. These factors may be technological changes poor marketing efforts, changing consumer preferences etc. $n this case, the researcher may identify two factors and then convert them into a hypotheses that is subse uently tested by conclusive research .The process can be illustrated through the following figure 2.11 D&si1$ O. E<p,# !"# 3 S"+%i&s

Fxploratory studies are characteri)ed by their flexibility and ingenuity. Researcher use their imaginative skill in exploratory research to develop new ideas. Their research will be based on certain hypothesis. Eut as they proceed ,they may redefine their approach or they may proceed with a new set of hypothesis. 1o the researcher5s expertise is of paramount importance in exploratory studies The researcher can take three approaches to arrive at a meaningful hypothesis. '&

C 1tudy of secondary sources A 1urvey of individual who are expert in the subject B @ase analysis 2ague problem

8!ploratory 5esearch

&ypothesis

-ew .dea

7onclusive 5esearch

/ecision Fi1. 2.1 1ource2 -arketing Research, 9arper 3. Eoyd, Ir., Ralph 3estfall and 1tanley <. 1tasch. 2.12 S"+%3 #. S&)#$%! 3 D!"!

Researcher can utili)e the data compiled by other organi)ations to formulate the hypothesis. Research reports prepared by consultancies and marketing research organi)ations, sales data brought out by trade associates, survey reports of governmental and nongovernmental organi)ations are generally used for this purpose .1econdary data has proved to be uickest and most economical source for researcher. The information technology boom has made the search for data very easy .The internet is one of the largest repositions of secondary data, available at minimal or at no cost at all. 1everal $nformation Technology techni ues such as datamining help researcher collect the right data and also aid them in establishing connections. 6t times researcher may be confused by the information glut. $n such cases, the researcher should be prudent enough to select the relevant data and discard the rest. 6 detailed discussion regarding secondary data is given in the chapter 1econdary /ata. 2.13 S+ '&3 #. I$%i'i%+!,s 5i"( I%&!s

''

Researchers usually interview individuals who have a general idea about the subject and are also imaginative .They provide the necessary direction for researcher to focus on .1ales managers, sales representative and dealers can be interviewed for eliciting information. Fven the consumers are approached directly by different organi)ations for new ideas. <or instance, companies like A- has used this techni ue in the past. 3hile approaching consumers care should be taken to interview a heterogeneous group. $t is absolutely important for the researcher to find new ideas. 9ence research is usually conducted by interviewing people who are cooperative and have an interest in the subject being researched. Respondents should be given the freedom to express their ideas however radical they may be. 0ualitative research techni ue are used to collect exploratory data from individuals.These involve interviews with individuals and groups ./epth interviews or projectives techni ues are used to elicit information from individuals, while focus group interviews are employed to elicit information from groups. 2.14 D&p"( I$"& 'i&5s /epth interviews are conducted to elicit information (from consumers) that is difficult to obtain through direct interviews. <actors such as consumer attitudes and motivation are understood mainly through depth interviews. $n such interviews the researcher approaches the consumer with only an outline in mind. <ormal uestionnaires are not used in this techni ue. The interviewer may probe deeply to prompt the respondent to elaborate on new ideas. This is necessary because a direct uestion regarding the motivating factors behind a purchase may fail to elicit the appropriate answer. <or instance, a consumer may avail of five star hotel services, because he perceives such behavior as a statement of his social status .Eut he may not admit the factor that actually mot*ivates him. The researcher has to probe deep into the consumer5s mind by asking several indirect uestions. Crocter H Aamble conducts depth interviews to identify ambiguity in consumers5 answers and to understand the true meaning of their responses. $f a consumer says she uses a certain brand of shampoo because it does best job of getting her hair clean, researchers ask her, what does she mean by ,clean.8 /oes ,clean. mean the way it feels or the way it looks8 /oes ,clean. mean free of dirt or nongreasy or free of dandruff or a less itchy scalp8 $s ,clean. hair s ueaky, slippery, lively, bouncy, fluffy, shiny, easy!to!comb, or manageable8 7nly researchers with extensive experience and training will be successful in eliciting the information needed without any bias. 7ne major disadvantage of depth interviews is that their results cannot be compared as interviewers have different interview styles. 6nother major disadvantage is the difficulty in analysing the data. The data available is highly subjective and varies from one analyst to another. 2.16 P #8&)"i'& T&)($i*+&s '(

Crojective techni ues are used to provide extremely useful data regarding the attitudes and values of the respondent. These techni ues are based on the theory that the interpretation of any vague object or picture will reflect the individual5s background, attitudes and values. There are four categories of projective techni ues. construction and expression. Ass#)i!"i#$ T&)($i*+&s $n this techni ue the respondent is re uired to respond to the presentation of a stimulus with the first thing that comes into his mind. The respondents should give the first word that comes to their mind in a free word association techni ue while they should give a series of words or thoughts in a successive word association techni ue. This techni ue is normally used for testing potential brand names. $t is also used for measuring customer attitude about products or product attributes. C#2p,&"i#$ T&)($i*+&s $n this techni ue, the respondent is re uired to complete an incomplete stimulus. Researchers use two types of completion techni ues 2 sentence completion and story completion. This has proved to be an effective techni ue to understand more about a respondent5s attitudes and values. ?ormally, if a uestion is posed to respondents, they come up with an answer. Eut if they are asked to complete a sentence or a story, they will express a more revealing answer. <or instance, an incomplete sentence like, ,Ceople who like white rum are .......... will induce the respondents to come up with an answer that they genuinely feel is true. The story completion method is similar to the sentence completion techni ue. 9ere the respondents are asked to complete a story told to him. Msually, a specific situation, like a couple visiting a shopping mall and having a disagreement over the purchase of a product, is presented to the respondent for completion. $t has been found that the respondents will build the story using their own experience and attitudes. C#$s" +)"i#$ T&)($i*+&s @onstruction techni ues are uite similar to completion techni ues. These techni ues re uire the respondent to construct a story, dialogue or description. $n the cartoon techni ue the respondent may be asked to fill in the dialogues in a cartoon. $n the picture response techni ue, a picture will be shown to the respondents and they will be asked to interpret it. The picture will be vague so that the respondent has to use his imagination to interpret the picture. 2.1@ F#)+s 1 #+p i$"& 'i&5s '* They are association, completion,

Researchers conduct focus group interviews by bringing together consumers who have a common interest. The group will be interviewed by a researcher (who acts as the moderator). 6 large number of organi)ations now employ focus group interviews to gain more insights into customer preferences and expectations. <ocus group interviews provide ualitative inputs and normally do not measure the subject ualitatively. <or instance, Crocter H Aamble uses this techni ue ,to develop hypotheses for further exploration. or ,to help design. a uantitative study. $t uses a lot of focus groups to gain insights ! such as to explore whether using any fabric softener has any perceived connection with being a good mom or what kinds of reactions consumers might have to a new way of demonstrating Eounty paper towel5s absorbency. Eut, it doesn5t make road generali)ations and major decisions based on three or four focus groupsU. 7f late, research organi)ations have started recording the proceedings in order to do a detailed analysis later. $n a focus group interview, the moderator normally briefs the group about the topic to be discussed. Then the moderator throws out some uestions to the group. These uestions will usually be simple, often aimed at breaking the ice. <or instance, they may ask uestions like, 4what do you think about the product85 1uch uestions are easy to answer. This will slowly help generate a discussion among the group. 7nce the atmosphere is relaxed, the moderator may bring up more specific issues and carefully watch the proceedings so as to check whether the group is coming up with new ideas. Towards the end of the discussion the moderator may give the group a task. The moderator then leaves the room and watches the proceedings through a television (or a one way window) to see if the discussion has caused the client to think of any more uestions to ask. ?ormally the researcher will ask around nine to twelve uestions. The moderator also informs the group that the proceedings are being watched by another group (clients"researchers). Msually, the discussion is watched by the organi)ation5s staff through a one way window. 1pecial care has to be taken to see that the moderator blends with the group. $f the moderator is of the same age group and sex, the group members will express themselves freely. ?ormally a group consists of * to #% people. 9owever, groups can range from one to a do)en. This depends on the si)e of study being conducted. -embers for the group are selected on the basis of their familiarity with the product. 6nd if they are also articulate they will contribute more effectively to study. The sample should be dominated by the segment important for the project. $t has been found out that different forms of group for different segments yield good results. C S3$& 1is2 2 The combined effect of the group will produce a wider range of information, insight, and ideas than will the cumulation of the responses of a number of individuals when these replies re secured privately. A S$#5-!,,i$1 2 6 bandwagon effect often operates in a group interview situation in that a comment by one individual often triggers chain of responses from the other participants. '+

B S"i2+,!"i#$ 2 Msually after a brief introductory period the respondents get ,turned on. in that they want to express their ideas and expose their feelings as the general level or excitement over the topic increases in the group. ? S&)+ i"3 2 The participants can usually find comfort in the group in that their feelings are not greatly different from other participants and they are more willing to express their ideas and feelings. E Sp#$"!$&i"3 2 1ince individuals aren5t re uired to answer any given uestion in a group interview, their responses can be more spontaneous and less conventional, and should provide a more accurate picture of their position on some issues. H S& &$%ipi"3 2 $t is more often the case in a group rather than individual interview that some idea will ,drop out of the blue.. # Sp&)i!,is!"i#$ 2 The group interview allows the use of more highly trained, but more expensive, interviewer since a number of individuals are being ,interviewed. simultaneously. I S)i&$"i.i) s) +"i$3 2 The group interview allows closer scrutiny of the data collection process in that several observers can witness the session and it can be recorded for later playback and analysis. D S" +)"+ & 2 The group interview affords more flexibility than the individual interview with regard to the topics covered and the depth with which they are treated. C@ Sp&&% 2 1ince a number of individuals are being interviewed at the same time, the group interviews speeds up the data collection and analysis process.

2.1:

C!s& A$!,3sis

The case method involves examining a single or multiple situations when an organi)ation is addressing a problem. The situation may involve factors that are interrelated. The organi)ation may find the case method to be of absolute value sine it involves an in!depth examination of the problem. <or instance, an organi)ation which has noticed some variation in the uality of the product manufactured may feel that this variation is primarily due to several factors that may include proper design, testing, manufacturing processes, labor constraints, etc. $n order to deal with the problem, they may take up a case that is similar to the situation at hand. The situation is analy)ed thoroughly, thus helping them to arrive t a hypothesis. 2.1:1 C!s& M&"(#% D&si1$ ':

Researchers use analogy as a method of analysis in cases. Through cases, researchers attempt to find @ommon features to all cases in a general group <eatures which are not common to all groups, but common to some subgroups <eatures uni ue to a specific case

<eatures that are common and those that are uncommon are analysed thoroughly to formulate hypothesis. Researchers should be careful in selecting the cases for analysis. A%'!$"!1&s #. C!s& M&"(#% A%'!$"!1&s @ases are studied comprehensively, taking into consideration all aspects. Mnlike statistical studies which involve abstracts from real situation, describes a real!time situation. case study

Box -.12: 3'4S TO S/..'SS$/& $O./S "%O/PS 0ocus groups can be an e""ective mar,eting research tool. >ut li,e tools they need to be used properly in order to provide meaning"ul results. The most success"ul "ocus groups include the "ollowing characteristics. +ppropriate research ob5ectives : 5obert >ohle, 4resident o" 0ocus on .ssues, at a St. 3ouis: based mar,eting, consulting and research company, says the primary purpose o" "ocus groups is to test and develop hypothesis. (0ocus groups help de"ine various customer population segments. They help companies ma,e better 6udgements), says >ohle.

'D

9illiam -ewbold ,supervisor o" mar,eting research at /etroit 8dison adds, (0ocus groups are ideal "or concept testing ,copy testing and preliminary advertising testing. )The "ocus group "ormat allows the moderator to change things on the "ly and retest it. N 9hen you need a real "ast turnaround and "ast input, N"ocus groups are appropriateN, -ewbold says. >ohle points out, however, that "ocus groups have a ma6or limitation : they only provide directional in"ormation. The central "igure in "ocus groups is the moderator, who guides and leads the discussion. This role is crucial to the overall success o" the groups. &owever, a good moderator must wal, a tightrope between as,ing questions and eliciting "eedbac, "rom all the respondents. NThe moderator has to be able to manage without leading (the respondents) and has to be able to control strong personalities in the groupN, -ewbold says. NA moderator is #@ percent o" what you get "rom a "ocus group. The moderator has to ma,e everyone "eel important, so they will tal,.N NJou need a good moderator who ,nows the issues but isnOt de"ensive : a moderator canOt be too close to the issues. .t should be a third partyN, says 5obert Sit,aus,as, director o" communications technology "or /etroit 8disonOs 25< system. >ohle agrees. NThe discussion guide and the moderator are ,ey. The most important part is the ability o" the moderator to listen and probe without passing 6udgement.N "ood recruitin : Another ,ey to good "ocus groups is proper recruiting. 1ood representation is crucial "or achieving meaning"ul results. NThe recruiting should be really representative o" he customer baseN, says -ewbold. 5epresentative and balanced "ocus groups were one reason /etroit 8disonOs 25< groups were so success"ul. Sitaus,as says "ocus groups are ideal "or eliciting customer response "rom a variety o" demographic groups relatively quic,ly and easily. 0e** p*anned discussions uide : 9hile /etroit 8disonOs "ocus groups had a clear agenda, /etroit 8dison was care"ul to build "le!ibility and "luidity into the groups. NJou should have an genda, Sitaus,as e!plains, Nbut not a rule :based agenda.N >ohle adds that a discussion guider should be 6ust that : a guide. 4art o" the success will depend upon the ability o" the moderator and the respondents to go beyond the original guide and delve into the important underlying issues. .ndeed, the accessibility issue was never a part o" the original "ocus groups moderators guide. The utility thought power outages were the problem. &owever, the moderator uncovered inaccessibility as an underlying problem. Proper environment : 7reating the proper environment is another ,ey to the overall success o" "ocus groups. To be truly e""ective, the research sponsors must establish the proper setting. NThe setting has to provide the ,ind o" environment where you can communicate not what you want to hear.N To be truly e""ective, the research sponsors must establish the proper groups. N The setting hs to provide the ,ind o" environment where you can communicate not what you want to hear, but what you ought to hear,N >ohle says. !nterpretation : 0ocus groups are meaningless i" the "indings are not interpreted correctly. Jou need someone insight"ul to draw the conclusion "rom the groups. -ewbold says N4eople can 6ump to conclusions based on "ocus groups and can be misled by one strong personality. 9e advocate holding multiple groups.N Since the researchers will be in association with the respondent "or a longer period, they will develop an in"ormal relationship. This relationship will help in collecting more data. *oreover, the data available will be accurate.

(=

1ome of the disadvantages of case study are 2 @ase analysis is very subjective. $t is difficult to have a formal research method. Researchers tend to generali)e the situations, though the case may not call for such generali)ations. 2.2 CONCLUSIVE RESEARCH /DESCRIPTIVE RESEARCH OR EBPERIMENTAL RESEARCH0

@onclusive research provides information that helps the manager to evaluate and select a course of action. The decision!maker will have to choose one course of action from different alternatives. @onclusive research provides the relevant information to help the manager arrive at a decision. @onclusive research design is characteri)ed by formal research procedure. Research 7bjectives are clearly stated and information needs are explicitly stated in this type of research. 6 formal study procedure achieves a variety of research objectives 2 description of phenomena or characteristics associated with a subject population, estimation of the proportion of a population that have these characteristics, discovery of association among different variables and discovery and measurement of cause!and!effect relationship among variables. @onclusive research ca be classified as either descriptive or experimental. 3e will start the discussion with descriptive studies. 2.21 D&s) ip"i'& R&s&! )(

Mnlike exploratory studies, descriptive studies are characteri)ed by a formal design and an accurate description of the problem. This helps in identifying the information re uired and ensures that it covers all the areas re uired. $t is imperative that the design of descriptive studies be such that it specifies the source of information and the data to be collected from those sources. This is done mainly to ensure the accuracy and the appropriateness of the information collected. $t is e ually necessary to prevent the collection of any unnecessary data associated with such research. There are basically two types of descriptive studies. #. @ase method %. 1tatistical method C!s& M&"(#% This method is not often used in descriptive research. $t is more widely used in exploratory research. The procedure is the same as in exploratory research. The only difference between the two lies in the fact that while exploratory research offers flexibility, descriptive research is more structured, clearly defining the research problem and the points to be investigated. (#

S"!"is"i)!, M&"(#% This is the most widely used method in -arketing Research. $t makes use of techni ues that vary from simple means and percentages to very sophisticated techni ues. $n this method, a limited number of factors from a large number of cases are studied in!depth. Us& #. s"!"is"i)!, 2&"(#% 1tatistical tools are used by most marketing research professionals to understand the dynamics of the market. /ata is usually collected through observation or through interviewing. 1urveys to find the consumption patterns of children ad generate profiles of the people using the internet to make purchases are examples of the statistical method. ;et us see how <orrester research, the leading global marketing research agency, has studied online purchases. They have discovered that the average online buyer reports purchases in two product categories 2 books and software. The following are some of the observations made by <orrester on line purchases. -edia and technology lead net shopping. -ore than (=> of online shoppers buy software and more than '=> buy books. Euyers spend an average of more than V#== over three! month period. -ore than %(=,=== ?orth 6mericans bought cars and computers online. Three percent of online consumers report online securities trades.

A%'!$"!1&s #. S"!"is"i)!, M&"(#% 1tatistical study involves a large number of interviews or observation. 1tatistical techni ues are used specifically for mass data. Two different researchers conducting statistical research will arrive at the same results, while two people using the case study method may not arrive at the same results. This is mainly because of the subjectivity of the case study method. 1tatistical study helps the researcher to make more accurate generali)ations. $f the sampling is properly done, the generali)ation will be universally true. Dis!%'!$"!1&s <ails to prove cause!and!effect relationship /irection of causal effect is not clear in statistical studies. C!+s!, R&s&! )( @ausal research is also a techni ue used to perform conclusive research. $t attempts to specify the nature of the functional relationship between two or more variables in the problem model. -anagers analy)e the impact of advertising etc. through causal research. <or instance, once (%

they give an advertisement they study how the advertising has caused the sales to change. Msually conclusions are drawn from three types of evidence. @oncomitant variations 1e uence of concurrence 6bsence of other potential casual factors.

C#$)#2i"!$" V! i!"i#$s 6dvertising expenditures for an organi)ation vary across geographic regions. 6n organi)ation may notice the variations of sales across the region. $t may also notice that sales revenue is high in those regions where the advertising expenditures are high and is low in those regions where advertisement expenditures are low. 1o it may infer that sales revenue is directly proportional to advertisement expenditure. Eut this has only been inferred, not proved. S&*+&$)& #. O))+ &$)& This is another type of evidence that can be used to make inferences about causation. 1ome events that happen first, cause the next event to occur. $n such cases, researchers can infer that the first event has caused the second event. A-s&$)& #. O"(& P#"&$"i!, F!)"# s $f the causative factors are identified the researcher will be able to eliminate all the factors except the one that he believes is the real causative factor. Then he can establish the same as the real causative factor. 0uite often it is difficult to eliminate all the possible causative factors. 2.22 E<p& i2&$"!"i#$

Fxperimentation is another method used for conclusive research. $t can be used to find the cause and effect relationship between two or more variables. This is usually done by manipulating one or more variables (known as the dependent variable). Mnlike observational studies, the researcher systematically alters the variables of interest and observes the changes that follow. There are two types of experimentation, laboratory and field experiments. ;aboratory experiments are more controlled but are conducted in an artificial environment. The environment is totally man!made. <ield experiments are done in a natural environment and so are less controlled, but yield more realistic results. A%'!$"!1&s The degree to which the certainty of the causal relationship between two variables can be established is highest in experimental studies. This is mainly because the experimenter can (&

manipulate the independent variable and directly observe the effects of this manipulation on the dependent variable. 6t least in theory, the ability to manipulate is unlimited and the relationship can be established with complete certainty. The use of a control group is useful to assess the existence and potency of manipulation. $n experimental designs, one can isolate the experimental variables and thus avoid the effect of the extraneous variables on the results. The presence of control group assists further, due to the possible comparisons and the exclusion of the effects of the extraneous variables. Fxperimentation is a highly convenient method compared to others since the experimenter can adjust variables so as to achieve the extremes which are not observable under routine conditions, i.e. the manipulative control on the independent variable is high. -oreover, the researcher need not wait for the fortuitous occurrence of a particular incident to measure its effect, but can manipulate the variables to achieve the same. The experiments if repeated with various groups and various conditions can give rise to generali)ed theories about the relationship between the given variables. 7nce a theory is established, the need for many similar experiments is reduced. Researchers can use field experiments, to minimi)e distortions due to the intervention of the researcher on the results. 1uch experimentation also generally needs less financial resources than other methods. Dis!%'!$"!1&s $n laboratory experiments, artificiality is the main disadvantage. There are many internal and external validity problems uni ue to each experimental design. These are discussed in the following sections. Fxperimentation is not possible of the past and predictions based on experimentation are not possible for the distant future. 1o, it is applicable only to current problems or problems of the near future. The most important disadvantage is that the marketing research concerns itself with people and their behaviour, and so does not yield itself for thorough manipulation and ethical considerations limit these further. Fxperimentation is a familiar techni ue used by one and all in their daily lives. Eut it has been structured, theori)ed and systematically practiced by the scientific community to develop scientific knowledge and use it for social benefits. $n this techni ue, one identifies the change that takes place in a variable because of changes in other variables under controlled conditions. Though ancient physical scientists like 6rchimedes used scientific methods, the behavioural sciences did not take to experimentation in its systematic form until recently. This is due to the fact that it is not possible for behavioral scientists to control all the variables of the experimental environment to the extent it can be done in the physical sciences. 9owever this techni ue has recently gained popularity particularly due to the success of the techni ue in theori)ing the learning theories. ('

$n chemistry, for example, the scientist adds two chemicals, under controlled temperature, pressure and other variables, to find the effect of these on the reaction. $n marketing research, the researcher observes the purchase behavior of a customer, while controlling the variables like the amount of exposure to the promotion, store layouts, etc. The difficulties in applying the experimentation in marketing research lies in the fact that experimentation needs control of variables, and it is not easy to control variables in markets. The difficulties in doing so have been explained in the chapter ,1cientific -ethod.. $n the chapter ,Types of Research., we read about the various advantages and disadvantages of experimentation in comparison with other techni ues. This chapter deals with various types of experiments based on the settings namely, the laboratory and field settings. To make the discussion easier, it also defines the terminology and symbols used in the latter section. <urther, the potential threats to the validity of an experiment are discussed and various designs of experiments are also discussed in detail with their advantages and disadvantages. 2.: SOME ADDITIONAL TYPES OF EBPERIMENTAL RESEARCH DESIGNS 6) 6fter only design 2 This design consists of measuring dependent variable after exposing independent variable at Test units. E) I$p+" T&s" U$i"s O+"p+" Fxplanatory Bariables 1amples and /ependent Bariables Territories $ndependent variables like These are 1ales, ad! Croduct, Crice, Clace and recall, attitude, market Cromotion share etc. Fxtraneous Bariables Mn!@ontrollable Bariables ;ike competition E<!2p,&s ; a) -arketer distributes discount coupon to the consumers to buy the brand. The study measures the extent of coupons redeemed by the sample covered. Eritania5s campaign, 4britania khao, world cup jao5, involved scratch coupon to be surrendered to the retailer for matching the number. b) Reliance $ndustry ;td.,, in the month of Iune!Iuly every year disdtributes dividend warrants, along with which it also sends to shareholders %=> discount coupons for buying Bimal <abric. $n the month of <ebruary it conducts research to study number of coupons redeemed by the shareholders. c) Eefore!6fter /esign 2 $n this design the dependent variables are measured across test units with specific independent variable, once before the independent variable is ((

exposed to test units and again after exposure. The difference between the two measures is treated as the effect of experimental variable. E<!2p,& 2 Cepsi @o. measured the sales before launch of 47ya babli5 campaign and after the ad campaign. $t noticed healthy growth in the sales due to classic picturisation and content of the ad. @) B&.# &9A."& 5i"( C#$" #, G #+p 2 $n this design the research study includes a control group in the experiment, but the control group is not subjected to the experimentation. 1uppose the marketer wants to study the movement of market share with reference to price reduction, then the impact of experimentation is studied as follows 2 M&!s+ &2&$"s Eefore Fxperiment 6fter Fxperiment E<p& i2&$"!, G #+p M! 4&" S(! & F# F% C#$" #, G #+p M! 4&" S(! & @# @%

-athematically, impact of experimentation P W(F% Q F#) Q (@% Q @#)X E<!2p,& 2 -ost of the telecom companies reduce the tariffs, initially only for say one circle, study the impact of reduction in tariff on sales and market share and then repeat same strategy at other circles too. $n such cases, the circle where benefit is offered is called Fxperimental Aroup, whereas the consumer outside the circle is called @ontrol Aroup. They do not participate in experiment, but hope to get the benefit at later stage and hence buy the pre!paid or post!paid of same company which offered benefit at other circle. 2.:1 E<p& i2&$"!, R&s&! )( D&si1$ C!s& S"+%3 C!s& S"+%3 ; E<p& i2&$"!, R&s&! )( D&si1$ S . N#. #. N!2& #. C#2p!$3 6THT B !$% $dea Ti2& p& i#% #. I2p!)" &<p& i2&$"!"i#$ ;ife long recharge #st Ian. %==* to Three lacs new Rs.DD(, local calls &#st Ian. %==* consumers added Rs.#.DD, 1T/ calls Rs.%.DD. $n case of Cost! paid all calls 6t Rs.=.DD (#) Euy Reliance hand!set #st /ec. to&#st Ten lacs new at Rs.%(== and get /ec.%==( consumers added. e uivalent free talk time. (Total @onsumers #+= lacs as on (* D&"!i,s #. E<p& i2&$"

%.

Reliance

Reliance $ndia -obile

(%) Recharge forever

#.#.%==*. # Ian. to&# Ian. <ive lacs new %==* consumers added
st st

&. '.

E96R6T$ TF;F@79;;

6irtel ;iril

YRs.#,===, call any! 3here in $ndia. Re.#1-1 within the state 6t Re.=.=# ;ife!time recharge Rs.DDD #st Ian to&#st Ian. Three lacs sixty %==* thousand new consumers added a) @hange of ad!agency ?ovember%==( ?o change in from ;owe $ndia to -c brand sales @ann Frickson. from freshness, youth H exuberance to a young couple in a naughty mood with a slow humming jingle 4l!ee!ra! ee!ra5, exhibiting husband wanting to catch the wife nearby bathroom.

(+

CHAPTER 3 SOURCES OF SECONDARY DATA 3.0 INTRODUCTION

7nce the research process starts, the researcher charts out the research objectives. Then the researcher turns his attention to the research design and determines the sources of marketing data. 6t times, researchers make the mistake of conducting primary research for collecting data while data is available from secondary sources at a lower cost. -oreover, the secondary data sources provide information that may not be obtained by other external agencies. 9owever, the researcher should evaluate the secondary data for its reliability and accuracy. The researcher should also check whether the available data will fulfill the re uirements. $n this chapter we will discuss the nature of secondary data and its advantages and disadvantages. 3e shall also survey the sources of secondary data. 3.1 THE NATURE OF SECONDARY DATA

1econdary data is available from publications, in!house databases, research agencies etc. $t constitutes readymade information that can be used for research purpose with minimal analysis. 9owever, the researcher should bear in mind that secondary data is published for purposes other than the current research. @ollecting primary data involves field work and further analysis on the data collected to arrive at a conclusion. <or instance, a marketer who wants to launch a particular product may be interested in collecting data regarding the buying habits of consumers in that particular region. The marketer can conduct field surveys to collect the relevant data, which, in turn, can be analy)ed to arrive at a proper conclusion. Eut at the same time, he can refer to any published material that has already done an analysis. 3hile the first method is tedious, time consuming, and expensive, the second method, which is collecting secondary data, is fast and inexpensive. 3.2 ADVANTAGES OF SECONDARY DATA

7ne of the main advantages of secondary data is that it is uite inexpensive. 6 small start!up company study the market to launch a product may not be able to afford to do primary research. Ey getting hold of good reports and articles, such small organi)ations will be able to do the study cost effectively. 1econdary data helps researchers save time. 3hile primary research takes a considerable amount of time in the form of collecting and analy)ing the data, secondary data offers readymade solutions. $f the demographics of a particular region have to be studied, the researcher has to collect the statistics of the population. $t is impossible for any organi)ation to conduct such a census study. 9ere too, secondary data published by a government organi)ation will be of (:

considerable use. -oreover, data collected and published by the government will be less biased. 3.3 DISADVANTAGES OF SECONDARY DATA

The major disadvantages of secondary data are Relevance 6ccuracy 1ufficiency 6vailability 3.6 TYPES OF SECONDARY DATA

I$"& $!, !$% E<"& $!, D!"! The data available within the organi)ation, which may be published for purposes other than the problem at hand, is called the internal data. $nternal /ata may be sales reports, accounting records, inventory reports, budgets, profit and loss statement, etc. Fxternal data is the data available outside the organi)ation. This can be data made available to the organi)ation by external research organi)ations. 1yndicated sources publish and sell data periodically and library sources include information from a wide array of publications. C&$s+s D!"! $n the M1, the Eureau of @ensus publishes census data. /ata published by the Eureau of @ensus is known for its uality and credibility. $n $ndia, the census report is published by Registrar Aeneral of $ndia. The survey is usually done once in ten years. @ensus data contains demographic information. The report contains information on different aspects such as age, sex, education and occupation. Eut one of the main disadvantages of census data is that it has a time lag of three to four years. Net this is the only demographic data recogni)ed by users as authentic. C#22& )i!, I$.# 2!"i#$ 7rgani)ations periodically provide information. This is also referred to as syndicated research. This service is mainly aimed at meeting the re uirements of their clients. $nformation may be provided on a daily"weekly"yearly basis. Msually organi)ations providing such information conduct primary research to collect the data. 1ome organi)ations collect data from a sample growth over a period of time. This is called panel research. $t helps track changes in customer preferences over a period of time. -arketing Research organi)ations like 6@?eilsen, <orrester Research, Aallup -E6, 7RA, -6RA, $-RE, etc. sell information for a price. They usually study the client5s needs and customi)e their reports to suit those needs. $n $ndia, uasi!governmental institutions like the ?ational @ouncil of 6pplied Fconomic Research (?@6FR) and $ndira Aandhi $nstitute of (D

/evelopment and Research ($A$/R) sells market information and is considered to be highly ualitative. C#$s+2& P+ )(!s& D!"! @onsumer purchase data is extremely useful for devising marketing strategies. $t provides information which can be used for understanding the market share, market segments, competitors, effects of advertising, etc. 1yndicated Research agencies provide such information on a continual basis. R&"!i, !$% A(#,&s!,& S!,&s D!"! 6gencies like 7RA!-6RA, <rancis Ganoi and 6@?eilsen conduct studies and provide information on sales data for a number of items. These studies are mainly on packaged foods, toiletries, cosmetics and over!the!counter drugs. /ifferent agencies collect data at different intervals. They usually collect data from stores located in different towns. A%'& "isi$1 D!"! 7rgani)ations spend a huge amount of money on advertising every year. 7bviously, they will be interested in knowing the readership of various newspapers and maga)ines. They will also be interested in understanding the television viewing patterns. The ?ational Readership 1urvey (?R1) is conducted in $ndia to assess the consumer profile of newspapers and maga)ines. ?R1 is conducted by the ?ational Readership 1tudies @ouncil. $t is constituted of the 6dvertising 6ssociation of $ndia, the 6udit Eureau of @ir(cultion and the $ndian ?ewspaper 1ociety ($?1). The actual fieldwork is carried out by 6@?eilsen, $-RE, and Taylor ?elson 1ofres -ode. $n $ndia, $-RE conducts studies and publishes data on television viewing patterns. 6ll this information is used by or'ganisations for media planning. T&s" M! 4&"i$1 1ome marketing research agencies provide test!marketing services. 6 new product, before it is launched in the market, is introduced in a test market and its availability is ensured. Then the sales data is collected on a periodical basis and reported to the client. 3.@ 3.@1 ADDITIONAL SOURCES OF SECONDARY DATA G#'& $2&$" S#+ )&s I$.# 2!"i#$ p #'i%&% $nstalled manufacturing capacities H actual utili)ed capacities for all manufacturers 6vailability of foreign currencies. $mport!Fxport statistics

N!2& #. "(& S#+ )& #) /irectorate Aeneral of 1upplies H /isposal (/A1H/) %) /irectorate Aeneral of Trade H /isposal &) Reserve Eank of $ndia (RE$) ') /irectorate Aeneral of @ommercial *=

$ntelligence H 1tatistics () @entre for monitoring $ndian Fconomy (@-$F) *) @ensus +) Aeographic 1urvey of $ndia :) 9orticulture Eoard of $ndia

Fconomic Arowth, A/C Copulation, no. of families, no.of voters Regionwise production of agri!produce Balue!added fruits, vegetables H flowers and markets

*#

N!2& #. "(& S#+ )& D) /irectorate Aeneral of <oreign Trade (/A<T) #=) Fxim Eank ##) Fxport @redit Auarantee @orporation of $ndia (F@A@) #%) 6griculture H Crocessed <ood Fxport /evelopment 6uthority(6CF/6) #&) @entral 1tatistical 7rganisation (@17) #') ?ational 1ample 1urvey (?11)

I$.# 2!"i#$ p #'i%&% $mport Fxport Regulations @reditworthiness of importers and countries. $nsurance covers and financial guarantees available to exporters. 9igh Tech 6gri <arming, technology tie! ups, seed capital, inspection, etc. $ndustry Fconomics Cer @apita consumption H monthly per capita income, literacy per state, employment across male H female etc.

3.@2

N#$9G#'& $2&$" S#+ )&s TRC ratings, Retail 1tore 6udit 6ny publication after #D+= 9ousehold disposable income H consumer behaviour. Barious courses, fees, duration and eligibility. @lassified information @lassified information ?o.of Retailers, their classification, types, etc. @ustomer satisfaction

#) 7rg -arg %) $?1/7@ (private ;ibrary) &) Cath <inder ') Mniversity Cublic Relation 7ffices () Nellow Cages H 6sk -e *) $nternet 1ites +) $ndian 6ssociation of Retailers :) I./.Cower 6sia Cacific

*%

CHAPTER 4 HYPOTHESIS 4.0 INTRODUCTION

$n the chapter on preparation and tabulation of data we discussed the appropriate procedures for collection and tabulation. 7nce we tabulate the data we need to analy)e it, i.e. is we should verify the hypothesis stated in the problem. To do so we need to learn hypothesis! testing methods. $f the manager of a shopping mall wants to find out if customer satisfaction is at least D= percent, we can test the validity of this hypothetical parameter by the use of hypothesis testing. 9ypotheses test, also known as tests of significance, enable us to decide on the basis of the sample results if the deviation between the observed sample statistic and the hypothetical parameter value (or) statistic is significant (or) might be attributed to chance (or) the fluctuations of sampling. 4.1 METHOD OF HYPOTHESIS TESTING

/efinitions of 9ypothesis (i) 9ypothesis Q $t is a statement or assertion about the statistical distributor or parameter of statistical distribution. 6lternatively hypothesis is a claim to be tested. ?ull hypothesis Q 6 hypothesis of 4no difference5 is called null hypothesis 6lternative 9ypothesis Q $t is a hypothesis to be accepted in case null hypothesis is rejected. $n other words, a complementary hypothesis to null hypothesis is called alternative hypothesis. S"&ps I$ F# 2+,!"i$1 A$% T&s"i$1

(ii) (iii)

4.12

Testing for statistical significance follows a well!defined pattern. Though one may not be able to understand all the terms in these steps at this stage, we are mentioning them here. They will be discussed in subse uent chapters. The steps are as follows2 1tate the null hypothesis2 The null hypothesis must be stated. @hoose the statistical test2 The choice of the statistical test is dependent on the power and efficiency of the test, the nature of the population, the method of drawing the sample and the type of measurement scale. 1elect the desired level of significance2 The exact level of choice depends on how much 6lpha risk one is willing to take in comparison with beta risk (6lpha risk and Eeta risk are explained later in this chapter). @ompute the calculated difference value2 6fter the data is collected, the formula for the appropriate significance test should be used to obtain the calculated value. *&

7btain critical test value2 The critical value for the calculated value should be looked up in the appropriate tables. The critical value is the criterion that defines the region of rejection from the region of acceptance of the null hypothesis. -ake the decision2 <or most tests, if the calculated value is larger than the critical value, we reject the null hypothesis rejected and it is conclude that the alternate hypothesis is accepted. $f the critical value is larger, we conclude we have failed to reject the null.

A.EK o" area 5e6ection region

DEK o" area A.EK o" area Acceptance region 5e6ection region

4.13

F# 2+,!"i$1 A H3p#"(&sis

The first in hypothesis testing is stating the hypothesis itself. 6 hypothesis to a problem can be basically stated in two ways Q ?ull hypothesis and 6lternative 9ypothesis. ?ull 9ypothesis2 $n tests of hypothesis we always begin with the assumption (or) hypothesis called ?ull 9ypothesis. The ?ull hypothesis asserts that there is no significant difference between the statistics and the population parametersJ and whatever observed difference is there is merely due to population. $t is denoted by the symbol 9=. The null hypothesis is often the reverse of what the experimenter actually believesJ it is put forward to allow the data bring out the contradiction. $n the above example, the null hypothesis is that the average purchase has not changed from Rs. #(==. it is represented by 9= 2 Z (mu) P Rs. #(== 6lternative 9ypothesis2 6lternative hypothesis is complementary to the ?ull hypothesis and is denoted by the symbol 9#. $n the above example, the alternative hypothesis is that there has been a change in the average purchases per week from Rs. #(==. 3e can have three different alternative hypotheses about this change. These are indicated below as2 *'

96 2 Z (mu) [ Rs #(== 962 Z (mu) \ Rs #(== 96 2 Z (mu)T Rs #(== The hypothesis can be tested with a two!tailed test. The regions of rejection for null hypothesis are divided between the two tails. The second hypothesis uses the right tail for rejecting the null hypothesis whereas the third uses the left tail for rejecting it. 6 hypothesis is never acceptedJ it is only rejected or failed to be rejected. This statistical testing is not sufficient proof for disproving a hypothesis. Eut instead of a clumsily saying that we have failed to reject the hypothesis, we say that we accept the hypothesis. Rejecting a null hypothesis is e uivalent to accepting the alternative hypothesis and rejecting an alternative hypothesis is e uivalent to accepting the null hypothesis. 4.14 E # s I$ T&s"i$1

The decision to accept or reject the null hypothesis 9= is made on the basis of the information supplied by the observed sample observations. The conclusion drawn on the basis of a particular sample may not always be true with respect to the population. <or instance, in the above mentioned example we have a (.=> chance of rejecting a true hypothesis in the above mentioned example. $n table '.#', four cases are presented. 3hen the alternative hypothesis is true, it means that the null hypothesis is false. Msing this concept we can deduce that the cases are accepting a true null hypothesis and rejecting a false null hypothesis from the table it is clear that in any testing problem we are liable to two types of errors. Type!$ error2 Rejecting a true null hypothesis is called a Type!$ error. $t is compared to convicting an innocent person. This is considered a serious error and researchers generally try to minimi)e its occurrence as much as possible. The probability of rejecting a true null hypothesis in the above example is (>. This indicates the probability of a type $ error. $t is denoted by ]. 9ere, ] P =.=(, or (> The region between the acceptance and rejection region is called the critical value. $n the above problem the critical values are Rs. #'+= and Rs. #(&= at a given significance level of (>. 6lternatively, for a given significance level we can calculate the critical values above or below which a hypothesis can be rejected or accepted. Type!$$ error2 6ccepting a false null hypothesis is called a Type $$ error and is compared to ac uitting a guilty person. $t is difficult to detect such an error. $t is denoted by ^. 6nd this *(

error depends on (#) the true value of the parameter, (%) the ] level we have selected, (&) the nature of the test used (one or two!tailed) to evaluate the hypothesis, (') the sample standard deviation, and (() the si)e of the example. ;et us assume that the mean has actually moved from #(== to #'+=. 7ur null hypothesis is that the average purchase is #(==. This is false. The probability of not finding this out, which is nothing but assuming that the given hypothesis is correct, is (^) D(>. <or a different population mean the value of ^ will be different. $deally, a )ero ^ indicates an error free test. This means that ideally #! ^ must be e ual to #. The closer this value is to #, the better is the test. #! ^ is considered as the power if a hypothesis test for it is the probability of rejecting a false null hypothesis. 9= is true 96 is true 4.16 A))&p" H0 @orrect 3rong Q Type!$$ error R&8&)" H0 3rong Q Type!$ error @orrect

S&,&)"i$1 A T&s"

Three uestions should be raised when choosing between various tests. 9ow many samples does the test involve8 7ne, two or G8 $f moor than one sample is involved, are they related or not8 3hat is the type of data8 ?ominal, ordinal, interval, or ratio8

0uestions like the si)e of the sample, the uality of the sample si)e and weighted data can be raised. These uestions will be answered in advanced statistics books and researchers should make use of them when re uired. Two samples are often used when there are two different products. Two samples, one for each product, are taken and tested to find out whether they belong to the same population. Table '.# lists the various statistical techni ues appropriate for different measurement levels and test situations. 6?7B6 is discussed in the text, but in a separate chapter. 7nly the most commonly used tests are surveyed in the following sections. ?on!parametric tests except chi s uare tests, call for an involved discussion and so are not discussed here. Refer to advanced spastics books for studying these methods in detail. 4.2 CHI SCUARE /D20 ANALYSIS

This is the most widely used non!parametric test, particularly for nominal data, but it can also be used for higher scales. $t is used for actual values rather than percentages. $t is used to find if difference between the
,

_ P`
i$C

(+i = 8i)A 8i **

observed distribution of data among categories and expected distribution is significant. O$& s!2p,& T&s" $n this test, we first note the expected (hypothesi)ed) fre uencies in each of the categories. Then the values of actual fre uencies are compared with the hypothesi)ed fre uencies. The value _% is a measure that expresses these differences in the form of a mathematical value. The larger this difference, the larger this difference, the larger is the _% value. The formula for _% is given as 3here 7i P 7bserved number of cases categori)ed in the ith category Fi P expected number of cases in the ith category GP The number of categories _% is uni ue for each degree of freedom. The degrees of freedom involved in a category are e ual to G!#. @are should be taken in using the chi s uare method in the following cases2 3hen d.f. P#,each expected fre uency should be at least ( in si)e. $f d.f.\#, then the _% test should not be used if more than %= percent of the expected fre uencies are smaller than (, or when any expected fre uency is less than #.

;et us take an example. 6 survey was conducted in /elhi to measure the intent of purchasing a second car. 6 sample of %== people was taken. 3e would like to analy)e the data based on the profession of the respondents. $s the intent dependent on the profession or not8 3e assume that these categories have no effect on the income. ?ow we proceed with the procedure recommended earlier. 9ypothesis 2 9=2 7# Q Fi. The proportion of the population that intends to buy independent of their professional categories as given. 6lternative hypothesis is 9627#T\ Fi 1tatistical test2 The responses are divided into nominal categories and so we should use @hi s uare analysis @alculated value2 Msing the Table '.% we have calculated the value chi!s uare to be _%P #%.*: *+

/egrees of freedom are '!#P& @ritical value2 <rom the tables we get a critical value of +.:% for a significance of (>. /ecision2 here the calculated value is greater than the critical value and so we reject the null hypothesis conclude that the categories do have an effect on the intent to purchase a new car. T!-,& 4.2; T(& D!"! !$% C!,)+,!"i#$s .# C(i9S*+! & 5i"( Si$1,& S!2p,& P #-,&2 P #.&ssi#$ I$"&$%&$" "# -+3 Oi #' #+ #' #( *= N+2-& i$"& 'i&5&% D= '= '= &= %== P& )&$" /N#. I$"& 'i&5&%? 2000 '( %= %= #( #== E<p&)"&% F &*+&$)i&s /P& )&$" < @00 Ei %+ #% #% D *= /Oi9Ei02

1elf employed (like doctors, lawyers) <ront ;ine workers 6dministrative 6cademic Total T5# S!2p,& T&s"

&.%* %.=: =.&& '.== #%.*:

The basic methodology is same as in the one sample test but the formula involved is as follows2 9ere the data is categori)ed and so is placed in a two
(+i6 : 8i6)

_ P``
i 6

8i6

/imensional matrix. The subscript ij refers to ijth cell. The degree of freedom are given as (r!#)(c!#). 4.3 ONE AND TAO TAILED HYPOTHESIS

There could be two types of situations, based on which hypothesis is classified as one sided or one tailed and tow sided or two tailed. 3hen alternate hypothesis 96 is defined as only more than or less than hypothesi)ed mean (Z) i.e. 96\ Z i.e. 96 \ Z or 96 T Z is called one tailed hypothesis. 7n the other side when alternate hypothesis is stated as not e ual to hypothesi)ed mean (Z) i.e. 9 6 [ Z, it means 96 could be less than Z or more than Z. 9ence this is called as two sided or two tailed hypothesis. 4.4 LEVEL OF SIGNIFICANCE AND CRITICAL VALUE OF E *:

L&'&, #. i1$# !$)& /F0 #=> (> #> 4.6 ILLUSTRATIONS

C i"i)!, '!,+& #. E O$& "!i,&% "&s" /GF0 T5# "!i,&% "&s" /GF0 #.%: #.*' #.*' #.D* %.&& %.(:

C!s& /i0 T5# "!i,&% "&s" Croblem 2 ?icrome -etal works, a leading name in Cackaging $ndustry, has designed automatic milk packing mache 4<ill!Cack5 to fill plastic pouch with # litre of milk with a standard deviation of =.=# litre. 6 sample of #== pouches was examined and then the average volume " uantity of milk found was =.D: litre. @an we say with D(> confidence that the machine is working property8 ?ull 9ypothesis P 9= P # lit. 6lternate 9ypothesis P 96 [ # ;it R !Z Test 1tatistics P t") P !!!!!!!!!!! 1"a n /ata 2 x P =.D: lit., Z P 9=P# lit, n P sample si)e P #==, standard deviation P s P =.=# lit. =.D: !# ! =.=% 9ence t") P !!!!!!!!!!!!!!!! P !!!!!!!!! P %= =.=#" #== =.=# <or D(> confidence level, corresponding level of significance is (>, and the value of ) for two tailed test is #.D*. 6s such calculated value of ) i.e. %= is more than actual value of ) i.e. #.D*. 9ence null hypothesis is rejected. @onclusion Q Cacking machine is not working properly. C!s& /ii0; O$& "!i,&% "&s" 6 sample of #=== spherical roller bearing is found to have average weight of (= grams. 1ample population standard deviation is ( gm. 7ne bearing, randomly selected was found of *= gm. 3hat is the guarantee that balance bearing will be of correct weight8 ?ull 9ypothesis P 9=PZ Q (= gm 6lternate 9ypothesis P 9#2 ZT (= gm x!Z Test statistics P t") P !!!!!!!!! 1 " an *D

/ata2 xP*= gm, Z P (=gm, 1P( gm, n P sample si)e P#=== *=!(= #= #= P !!!!!!!!!!! P !!!!!!!!! P !!!!!!!!!!! P *&.%D (a#=== ( " &#.*% =.#(:

t")

6ssume level of significance #> hence value of ) for one tailed test is %.&&. 1ince calculated value of b (*&.%D) is much more than actual value of ) (%.&&) null hypothesis is rejected. @onclusion2 There is no guarantee that remaining bearings will be of correct weight of (= gm. $llustration Q ;ux 1oap and changing consumer behaviour R&s&! )( P #-,&2 #) %) 3hether 1hah Rukh Ghan is the right choice as a male ambassador for ;ux. To test this we will have to find out whether people associate 1hah Rukh Ghan5s ualities with ;ux. 3e shall also analy)e whether ;ux needs to target the male consumers also. 3e shall test by finding out whether men really have a say in the purchase decision for soaps.

RF1F6R@9 7EIF@T$BF Crimary 7bjective To find whether there is an image mismatch between the image of 1hah Rukh Ghan and ;ux 1econdary 7bjective : : To find out whether the new improved positioning of ;ux (targeting men also) is re uired8 To find out which male celebrity (if any) is the most appropriate for ;ux.

RF1F6R@9 9NC7T9F1$1 9= 2 96 2 There is no mismatch between the image of ;ux and 1hah Rukh Ghan. Tehre is a mismatch between the image of ;ux and 1hah Rukh Ghan.

<$?/$?A1 +=

# % & ' ( *

-asculine <eminine 1tatus 1ophisticated @ool " 9ep Alamorous

0ualities ( # & ' ( %

Rank for ;ux * % ( & ' #

Rank for 1RG = # % # # #

1ummary of <indings # % & ' ( * ANALYSIS -arket 1hare of Barious 1oaps2 7ut fo the total %# men interviewed ( of them use ;ux, & use @inthol and & use dove. D of them use soaps other than those mentioned here. <rom among the +& women interviewd, #: of them use ;ux i.e. around %(> of them use ;ux, #&> use dove and pears and around #+> use @inthol. 6 very small share goes to /ettol, i.e. around %.+> whereas around %(> of the female respondents prefer to use other soaps like @handrika and other medicated soaps. -en and Euying /ecision2 <rom, the data collected, we have found that out fo all the men interviewed only %#> either buy the soap themselves or ask someone else to buy the brand they specify. This means that the men do not have any influence on the buying decision for any brand of soap. Thus the strategy of ;ux of trying to capture the male segment of the society by targeting them would not work as men do not influence the buying decision. Reason for buying a 1oap2 <rom what we have collected, we find that just '> of the respondents buy soap because a celebrity endorses it. -ajority of the respondents buy it for medical reasons (&#>) or because of its attractive packaging, shape or scent (%'>). ?ot even one of these respondents claims to be buying their brand of soap due to influence by friends or peers. 6 good #'> of the respondents buy their brand of soap because they think their brand is value for money. Thus we can infer that the buying decision for a particular brand of soap largely depends upon +# Alamour <eminist -ascutinity 1tatus 1ophistication @ool " 9ap @orrelation There is not a significant /ifference There is a significant /ifference There is a significant /ifference There is a significant /ifference There is not a significant /ifference There is a significant /ifference

medical reasons. Thus ;ux would be better off trying to capture more market share by using that strategy. ;ux and -ale @elebrity8 3hen the respondents were asked about their opinion on ;ux using a male celebrity to endorse it, only %&> said that they liked the idea, %=> said that it does not matter to them whether ;ux uses a male or a female celebrity, whereas (+> of them did not like the idea. This shows that the 6d has not been absolutely accepted by the general public. The 6d might have created a stir in the market, but that according to us will not attract many customers and may be not improve the market share for the soap. Creferred male 1tar in ;ux 6d2 $ fhte company plans to continue with a male celebrity in its ads, it should take 1aif 6li Ghan as majority of the respondents (&#>) though he is more of a metro sexual man as compared to 1RG. 9e was followed by 1hahid Gapoor with #D> votes. This supports our research as it shows that the public have not been able to connect to the ;ux!1hah Rukh Ghan partnership and the company would been better off if they would have chosen either 1aif 6li Ghan or 1hahid Gapoor for the same. Eoth 1aif and 1hahid are on the up in their careers and are thus very much in the news. The metro sexuality uotient is very high in both of them and thus very much liked by youth of today. ;MR 6/ <F6TMR$?A 1969 RMG9 G96?2 ?early # out of every % people asked did not like the ;ux 6dvertisement featuring 1hah Rukh Ghan. 6lso, the number of people who liked the 6dvertisement is a #D>. -ost of the people were of the opinion that the advertisement was not only unaesthetic but also that it could have been shot in a better manner as it was the first time that ;ux was experimenting with a male celebrity. The advertisement has created a stir in the minds of the consumers, but has not necessarily helped the company in increasing its market share. The advertisement is in the news, but not for the reasons the company must have wanted it to be in. this has lead to confusion in the minds of the consumers. TF1T$?A 7< 9NC7T9F1$1 To test our stated hypothesis, we wanted to see if there was a correlation between the ualities associated with ;ux and those associated with 1hah Rukh Ghan. There is significant difference in the rankings given to the ualities for each 1hah Rukh Ghan and ;ux. #. %. &. There is no significant /ifference in the Alamour uotient of 1hah Rukh Ghan and ;ux. There is a significance /ifference in the <eminine uotient of 1hah Rukh Ghan and ;ux. There is no significance /ifference in the 1opyhistication uotenti of 1hah Rukh Ghan and ;ux. +%

'. (.

There is a significance /ifference in the -asculinity uotient of 1hah Rukh Ghan and ;ux. There is a significance /ifference in the 1tatus uotient of 1hah Rukh Ghan and ;ux.

Thus, we find that there is a significant difference in the uotients of ' of the * ualities used to describe 1hah Rukh Ghan and ;ux. Thus we conclude that there is an image mismatch to some extent. Therefore, we reject the null hypothesis (that there is no mismatch between the image of ;ux and 1hah Rukh Ghan), thus concluding that, There is an image mismatch between ;ux and 1hah Rukh Ghan.

+&

CHAPTER 6 SAMPLING 6.0 INTRODUCTION

6n important step in the data collection process is sampling. 1ampling is the process of selecting a representative part of a population, studying it and thereby drawing conclusions about the population itself. The most commonplace examples of the sampling techni ue are tasting a small part of a dish to determine its taste, testing the temperature of water in a bathtub by dipping a finger, glancing through a book before buying it, etc. 1ampling is a very important aspect of marketing research and due care has to be taken to arrive at the right sample to be studied. 7ften it is impossible or too expensive to study the entire population for the decision!maker to understand the market. $n this chapter we will discuss the basic concepts of sampling, types of sample designs and calculation of sample si)e. 6.1 THE SAMPLING TERMINOLOGY

The terminology of sampling has evolved over the period of its existence. Gnowledge of these terms is necessary for understanding sampling. ;et us examine these terms through the hypothetical case of 3ild goose, a marketing research firm. This firm wants to find out the types of movies the owners of B@/ players in $ndia would like to watch. 6.11 E,&2&$"

6n element is a unit of study, which is measured for the purposes of research. This can be an individual or an organi)ation or even inanimate objects like soaps manufactured in a production line. $n the example mentioned above, the families owning B@/ players constitute the elements. 6.12 P#p+,!"i#$

The total collection of elements under investigation is known as the population. $n the study conducted by 3ild Aoose, all the families owning B@/ players form the population of the study or (a) 6ll members who buy branded baby products (b) all teenagers who watch -TB. 6.13 S!2p,&

The subset of the elements of the population chosen for study is called the sample or the study sample. The characteristics of a good sample are discussed later in the chapter. 3ild Aoose may choose a few cities for sampling and within these cities it may further select a few families. The list of the families, thus selected, forms the sample used in the study. 6.14 S!2p,i$1 U$i"s

1ampling units are non!overlapping elements from a population. 6 sampling unit can be an individual element or a set of elements based on the sampling process used. $f 3ild Aoose +'

uses simple random sampling, it considers a single family as a unit. $f it uses cluster sampling, it views each cluster of families as its primary sampling unit, while the individual family becomes the secondary!sampling unit. 6.16 S!2p,i$1 . !2&

The sampling frame refers to a complete enumeration"list of the population as specified by the research problem. $t is a list of all the sampling units. <or example, a list of all the people in the country owning B@/ players constitutes a sampling frame. 7ne should be careful in designing and selecting the sampling frame. 3ild Aoose may obtain its sampling frame from all the manufacturers of branded B@/ players, but this frame will not be exhaustive, as it will not include the people who made their purchase from the unorgani)ed sector. Fxample (a) Telephone directory of any city (b) ;ist of $ncome Tax Cayers. 6.1@ S!2p,i$1 E # s

There are two types of errors (i) imprecision inherent in using statistics to estimate parameters and, (ii) errors associated with applying a decided sampling procedure. $f probability samples are used, sampling theory can estimate the degree of imprecision that may be associated with a sampling design. 6.1: S!2p,i$1 P,!$

6 sampling plan is a formal method for specifying the sampling process of a particular study. 6.2 THE NEED FOR SAMPLING

$f one measures each and every element of a population for some characteristics of interest, the study is referred to as a census. Eut if one selects a small subset of the population for the study and then generali)es the results to the entire population, then it referred to as sampling. 1ampling is an attractive alternative to the census method as can be seen from the following discussion. 6.3 CHARACTERISTICS OF A GOOD SAMPLE 6 good sample should be accurate. 6ccuracy is a measure of the absence of bias or the absence of systematic variance. 1ystematic variation is the variation of a measure in one direction due to some known factors. The following example illustrates this definition. 6 good sample should be precise, that is, it should have a low standard error of its estimate. 6 good sample should be able to specify the accuracy and precision associated with it.

+(

$t should enable researchers to specify the degree of confidence that can be placed in its parameter estimate. SAMPLING DESIGNS

6.4

1ampling /esigns are of two major types2 probability and non!probability methods. These two types are presented in Table (.& along with five important considerations. Fach of these two types comprises a variety of methods of sampling. The characteristics given in the table are not accurate to describe each of the sub designs, but only give the overall characteristics of the probability and non!probability designs. Crobability samples are considered to be more costly because they need a sampling frame of the entire population. -oreover, in probability sampling, the selected sample units may be located at inconveniently distant places, thus entailing higher expenses for covering them. This type of sampling also takes more time because it is systematic. Eut this method is very accurate due to its known probability distribution. 1ince the properties of the sample are well defined and predictable, they are generally accepted. /ue to their strong theoretical base, these studies are replicable and their results are generalisable. Eefore we deal with complex sampling methods, we shall study basic sampling concepts. T!-,& 6.3 ; S!2p,i$1 D&si1$ C(#i)& C#$si%& !"i#$s C#$si%& !"i#$ @ost 6ccuracy Time 6cceptance of results Aeneralisability of results P #-!-i,i"3 -ore @ostly D&si1$ T3p& N#$9P #-!-i,i"3 ;ess @ostly ;ess 6ccurate ;ess Time Reasonable acceptance Coor

-ore 6ccurate -ore Time Mniversal acceptance Aood

T!-,& 6.@ ; S!2p,i$1 M&"(#%s # T3p&s #. S!2p,i$1 %&si1$s T3p& #. S!2p,i$1 Crobability designs 6. 1imple random B i&. D&s) ip"i#$ 6ssign to each #. population member a uni ue numberJ select sample items +* A%'!$"!1&s Re uires minimum #. knowledge of population in advance. Dis!%'!$"!1&s /oes to make use of knowledge of population that researcher may

by use of random %. numbers. &. b. 1ystematic Mse natural ordering or order populationJ select random starting point between# and the nearest integer to the sampling ratio (?"n)J select items at interval of nearest integer to sampling ration. @. -ultistage Mse a form of random random sampling in each of the sampling stages where there are at least two stages #.

<ree of possible classification errors. %. Fasy to analy)e data and compute errors $f population is #. ordered with respect to pertinent property, gives stratification effect and hence reduces variability compared to 6. %. 1implicity of drawing sample, easy to check

%.

have. ;arger errors for same sample si)e than in stratified sampling. $f sampling is related to periodic ordering of the population, increased variability may be introduced. estimates of error likely to be high where there is stratification

1ampling lists, #. Frrors likely to be identification and larger than in 6 or E numbering re uired for same sample only for members of si)e sampling units %. Frrors increase as selected in sample. number of sampling %. $f sampling units are units selected geographically decreases defined, cuts down field costs (i.e. travel) 3ith probability 1elect sampling reduces variability ;ack of knowledge of proportionate to units with si)e of each sampling si)e probability unit before selection proportionate to increases variability their si)e /. 1tratified 1elect from every #. 6ssures #. Re uires accurate #. Croportionate sampling unit at representativeness information on other than last with respect to proportion of stage, a random property that forms population in each sample basis of classifying stratumJ otherwise proportionate to unitsJ therefore, yields increases error. si)e of sampling less variability than 6 %. $f stratified lists are unit. or @, not available, may %. decreases chance of be costly to prepare failing to include themJ possibility of members of population faulty classification because of and hence increase classification process in variability ++

#.

characteristics of each stratum can be estimated and hence comparisons can be made. %. 7ptimum 1ame as /# except ;ess variability for same Re uires knowledge of allocation sample is sample si)e than /#. variability of pertinent proportionate to characteristic within variability within strata strata as well as their si)e. &. 1ame as /# except -ore efficient than /# for ;ess efficient than /# for /isproportionate that si)e of sample comparison of strata or determining population is not where different errors are characteristics i.e. more proportionate to optimum for different strata. variability for same si)e of sampling sample si)e. unit but is indicated by analytical considerations or convenience F. @luster 1elect sampling #. $f clusters are #. ;arger errors for units by some form geographically comparable si)e of random defined, yields lowest than other samplingJ ultimate field costs. probability samples. units are groupsJ %. Re uires only listing %. Re uires ability to select these at of individuals in assign each member random and take a selected clusters. of population complete count of &. @haracteristics of uni uely to a each. clusters as well as cluster, inability to those of population do so may result in can be estimated. duplication or '. @an be used for omission of subse uent samples, individuals since clusters, not individuals, are selected and substitution of individuals may be permissible <. 1tratified 1elect clusters at Reduces variability of plain #. /isadvantages of cluster random for every cluster sampling stratified sampling sampling unit. added to those of cluster sampling. %. 1ince cluster properties may +:

&.

A. Repetitive2 Two or more multiple or samples of any of se uential (doubt) the above types are taken, using results from earlier samples to design later ones or determine if they are necessary.

Crovides estimates of population characteristics that facilitate efficient planning of succeeding sampleJ therefore, reduces error or final estimate.

?on!probability /esigns Iudgment

0uota

1elect a subgroup of the population that, on the basis of available information, can be judged to be representative of the total populationJ take a complete count or sub sample of this group. @lassify population by per income propertiesJ determine desired proportion of sample from each classJ fix uotas

Reduces cost of preparing sample and fieldwork, since ultimate units can be selected so that they are close together.

change, advantage of stratification may be reduced and make sample unusable for later research. #. @omplicates administration of field work. %. -ore computation and analysis re uired than in non!repetitive sampling. &. 1e uential sampling can be used only where a very small sample can approximate representativeness and where the number of observations can be increased conveniently at any stage of the research. #. Bariability and bias of estimates cannot be measured or controlled. %. Re uires strong assumptions of considerable knowledge of population and subgroup selected. $ntroduces bias of observers5 classification of subjects and non random selection within classes

#. 1ame cost considerations as Iudgment (6dvantage) %. $ntroduces some stratification effect.

+D

#. @onvenience

%. 1nowball

for each observer. 1elect units of 0uick and inexpensive analysis in any convenient manner specified by the researcher 1elect units with 7nly highly specific rate characteristicsJ application. additional units are referred by initial respondents.

@ontains unknown amounts of both systematic and variable Representativeness of rare characteristic may not be apparent in sample selected.

6.:

ANOTHER APPROACH TO SAMPLE SIEE CALCULATION

(b 1) 1ample si)e P n P e where n P sample si)e, b P standard normal distribution for certain confidence level, 1 P population standard deviation and e P Tolerable error in estimating the variable. The value of 1 is calculated as follows 1 P population standard deviation P -aximum Balue Q -inimum Balue *

The denominator is * because DD.+> of the values of the variables would lie within c & x standard deviation i.e. & d I,,+s" !"i#$ 3hirlpool conducted @ustomer 1atisfaction 1urvey during /ecember %==( for 3ashing -achines. $t intended to measure customer satisfaction on a scale of # to #=, where # means not at all satisfied and #= means completely satisfied. 6ssume level of significance (> and tolerable error =.(. S#,+"i#$ ; <irst we compute 1 -ax Balue of @ust 1atisfaction Q -inimum Balue of @ust 1atisfaction 9ere 1 P * P #=!# * P #.( Balue of b for (> significance level is #.D( 9ence 1ample si)e P n P (b1)% F P (#.D( x #.()% =.( := P &(

6.7

SAMPLING TECHNICUES

P #-!-i,i"3 S!2p,i$1 N#$9p #-!-i,i"3 S!2p,i$1 #) Fach sampl'e unit in sample frame has #) The chance of each sample unit from sample e ual or know chance of being included as frame being included as sample cannot be sample estimated. %) 1amples are selected at random from sample %) 1amples are selected w.r.t. prior Fxperience frame. or judgement of the researcher &) 3henever large sample si)e is involved, this &) <or accessing small sample si)e this method is used. method is used. ') 3hen highly accurate decisions of known ') 3henever time and cost constraints are errors are intended regardless of cost, this inevitable (like exploratory Research), this method is useful. method is used. () ?ormally used for consumer goods survey. () ?ormally used for industrial goods survey. 6.= ILLUSTRATION Fmami wants to launch 4-adhuri5 and 4$shwarya5 range beauty creams, say in Cune. 9ow should it do sample design. S#,+"i#$; 1ample Copulation 2 6ll women in Cune. 1ample <rame 2 6ll women between age group #=!(= 1ampling -ethod 2 1tratified. 1ampling Clan , 1ample frame is divided into ' groups as follows 2 Aroup # Q 1chool!going girls between #=!#* Aroup % Q @ollege Qgoing girls between #+!%& Aroup & Q 3orking ladies between %' Q &( Aroup ' Q 9ousewives and working ladies between &*!(=. 1amples can be drawn from schools, colleges, offices, societies, etc. Iustification 2 Eeauty creams are costly and hence stratified sampling will ensure the income i.e. affordability. $t is seen that at higher secondary school level, the girls are more cautious about looks. 9ence, the age limit begins with #=. 6t the age (=, the ladies might value natural beauty. <our groups are formed to understand in depth the consumer profile and its preferences. 1ample si)e 2 #> from each group. (1ample frame for Cune contains : lacs ladies)

:#

CHAPTER @ MEASUREMENT IN RESEARCH @.0 INTRODUCTION

The task of measurement starts when the research problem is specified and a particular type of design is chosen. The importance of measurement should be understood by both the client and the researcher alike because it is the means by which reality is presented. 6 wrong measurement system or an incomprehensible measurement will make the research inade uate and sometimes invalid. $n short, the measurement should be well understood by the people associated with the research process, and the measurement system followed should be accurate and uniform. $n this chapter we will discuss the concept of measurement and the components of measurement. @.1 THE CONCEPT OF MEASUREMENT

-easurement can be defined as the assignment of numbers to the characteristics of objects, persons, states, or events according to certain rules. 6n object may be measured by its shape, length height, width or some other characteristics. 1imilarly, a study to determine whether children or adults consume a particular product more often measures the child!adult and purchaser!non!purchaser attributes of the persons sampled. The symbols used (referred to as numbers in the definition of measurement) depend on the nature of the characteristics they are to represent and how they are to represent them.

:%

CHAPTER : SCALING TECHNICUES FOR ATTITUDE MEASUREMENT :.0 INTRODUCTION

K6ttitudeK has been defined differently by many other authors. 1ome of the definitions are given below2 L6ttitudes are individual mental processes which determine both the actual and potential responses of each person in a social world. 1ince an attitude is alwaysK directed towards some object, it may be defined as Lthe state of mind of the individual towards a value.KL :.3 CLASSIFICATION OF SALES

The number assigning procedures or the scaling procedures may be broadly classified on one or more of the following bases2 (i) subject orientation (ii) response formJ (iii) degree of subjectivityJ (iv) scale propertiesJ (v), number of dimensions and (vi) scale construction techni ues. 3e take up each of these separately. (i) S+-8&)" # i&$"!"i#$2 Mnder it a scale may be designed to measure characteristics of the respondent who completes it or to judge the stimulus object which is presented to the respondent. $n respect of the former, we presume that the stimuli presented are sufficiently homogeneous so that the between! stimuli variation is small as compared to the variation among respondents. $n the later approach, we ask the respondent to judge some specific object in terms, of one or more dimensions and we presume mat the between!respondent variation will be small as compared to the variation among the different stimuli presented to respondents for judging. R&sp#$s& .# 22 Mnder this we may classify the scales as categorical and comparative. @ategorical scales are also known as rating scales. These scales are used when a respondent scores some object without direct reference to other objects. Mnder comparative scales, which are also known as ranking scales, the respondent is asked to compare two or more objects, $n this sense the respondent may state that one object is superior to the other or that three models of pen rank in order #, % and &. The essence of ranking is, in fact, a relative comparison of a certain property of two or more objects. D&1 && #. s+-8&)"i'i"32 3ith this basis the scale data may be based on whether we measure subjective personal preferences or simply make non!preference judgements. $n the former case, the respondent is asked to choose which person he favours or which solution he would like to see employed, whereas in the latter case he is simply asked to judge which person is more effective in some aspect or which solution will take fewer resources without reflecting any personal preference. S)!,& p #p& "i&s2 @onsidering scale properties, one may classify the scales as nominal, ordinal, interval and ratio scales. ?ominal scales merely classify without indicating order, distance or uni ue origin. 7rdinal scales indicate magnitude relationships of Kmore thanK or Kless thanK, but indicate no distance or uni ue origin. :&

(ii)

(iii)

(iv)

$nterval scales have both order and distance values, but no uni ue origin. Ratio scales possess all these features. (v) N+2-& #. %i2&$si#$s2 $n respect of this basis, scales can be classified as KunidimensionalK and KmultidimensionalK scales. Mnder the former we measure only one attribute of the respondent or object, whereas multidimension al scaling recogni)es that an object might be described better by using the concept of an attribute space of KnK dimensions, rather than a single!dimension continuum. S)!,& )#$s" +)"i#$ "&)($i*+&s2 <ollowing are the five main techni ues by which scales can be developed. (a) A -i" ! 3 !pp #!)(2 $t is an approach where scale is developed on ad hoc basis. This is the most widely used approach. $t is presumed that such scales measure the concepts for which they have been designed, although there is little evidence to support such an assumption. C#$s&$s+s !pp #!)(2 9ere a panel of judges evaluate the items chosen for inclusion in the instrument in terms of whether they are relevant to the topic area and unambiguous in implication. <or example Thurstone scale. I"&2 !$!,3sis !pp #!)(2 Mnder it a number of individual items are developed into a test which is given to a group of respondents. 6fter administering the test, the total scores are calculated for every one. $ndividual items are then analy)ed to determine which items discriminate between persons or objects with high total scores and those with low scores. <or example ;ikert 1cale. @umulative scales are chosen on the basis of their conforming to some ranking of items with ascending and descending discriminating power. <or instance, in such a scale the endorsement of an item representing an extreme position should also result in the endorsement of all items indicating a less extreme position. <actor scales may be constructed on the basis of intercorrelations of items which indicate that a common factor accounts for the relationship between items. This relationship is typically measured through factor analysis method. <or example 1emantic /ifferential scale.

(vi)

(b)

(c)

(d)

(e)

:.4

SCALING TECHNICUES.

3e now take up some of the important scaling techni ues often used in the context of research especially in context of social or business research. :.41 R!"i$1 s)!,&s2 The rating scale involves ualitative description of a limited number of aspects of a thing or of traits of a person. 3hen we use rating scales (or categorical scales), we judge an object in absolute terms against some specified criteria i.e., we judge properties of objects without reference to other similar objects. These ratings may be in such forms as Llike!dislikeL, Labove average, average, below averageL, or other classifications with more categories such as Llike very much!like some what!neutral!dislike somewhat!dislike very :'

muchLJ Lexcellent!good!average!below average!poorL, Lalways!often!occasionally!rarely! neverL, and so on. There is no specific rule whether to use a two!points scale, three!points scale or scale with still more points. $n practice, three to seven points scales are generally used for the simple reason that more points on a scale provide an opportunity for greater sensitivity of measurement. Rating scale may be either a graphic rating scale or an itemi)ed rating scale. :.411 The graphic rating scale is uite simple and is commonly used in practice. Mnder it the various points are usually put along the line to form a continuum and the rater indicates his rating by simply making a mark (such as ) at the appropriate point on a lineK that runs from one extreme to the other. 1cale!points with brief descriptions may be indicated along the line, their function being to assist the rater in performing his job. The following is an example of five!point graphic rating scale when we wish to ascertain peopleKs liking or disliking any product2 &ow do you li,e the productP (please chec,)

3i,e 2ery much

3i,e some what

-eutral

/isli,e some what

/isli,e very much

This type of scale has several limitations. The respondents may check at almost any position along the line which fact may increase the difficulty of analysis. The meanings of the terms like Lvery muchL and Lsome whatL may depend upon respondentKs frame of reference so much so that the statement might be challenged in terms of its e uivalency. 1everal other rating scale variants (e.g., boxes replacing line) may also be used. :.412 The itemi)ed rating scale (also known as numerical scale) presents a series of statements from which a respondent selects one as best reflecting his evaluation. These statements are ordered progressively in terms of more or less of some property. 6n example of itemi)ed scale can be given to illustrate it. 1uppose we wish to en uire as to how well does a worker get along with his fellow workers8 $n such a situation we may ask the respondent to select one, to express his opinion, from the following2 9e is almost always involved in some friction with a fellow worker. 9e is often at odds with one or more of his fellow workers. 9e sometimes gets involved in friction. 9e infre uently becomes involved in friction with others. 9e almost never gets involved in friction with fellow workers.

:(

The chief merit of this type of scale is that it provides more information and meaning to the rater, and thereby increases reliability. This form is relatively difficult to develop and the statements may not say exactly what the respondent would like to express. Rating scales have certain good points. The results obtained from their use compare favourably with alternative methods. They re uire less time, are interesting to use and have a wide range of applications. Eesides, they may also be used with a large number of properties or variables. Eut their value for measurement purposes depends upon the assumption that the respondents can and do make good judgements. $f the respondents are not very careful while rating, errors may occur. Three types of errors are common vi)., the error of leniency, the error of central tendency and the error of hallo effect. The error of leniency occurs when certain respondents are either easy raters or hard raters. 3hen raters are reluctant to give extreme judgements, the result is the error of central tendency. The error of hallo effect or the systematic bias occurs when the rater carries over a generali)ed impression of the subject from one rating to another. This sort of error takes place when we conclude for example, that a particular report is good because we like its form or that someone is intelligent because he agrees with us or has a pleasing personality. $n other words, hallo effect is likely to appear when the rater is asked to rate many factors, on a number of which he has no evidence for judgement. :.42 R!$4i$1 s)!,&s2 Mnder ranking scales (or comparative scales) we make relative judgements against other similar objects. The respondents under this method directly compare two or more objects and make choices among them. There are two generally used approaches of ranking scales vi). (a) -ethod of paired comparisons2 Mnder it the respondent can express his attitude by making a choice between two objects, say between a new flavour of soft drink and an established brand of drink. Eut when there are more than two stimuli to judge, the number of judgements re uired in a paired comparison is given by the formula2 n(n!#) ? P !!!!!!!!!!!! % where ? P number of judgements n P number of stimuli or objects to be judged. <or instance, if there are ten suggestions for bargaining proposals available to a workers union, there are '( paired comparisons that can be made with them. 3hen ? happens to be a big figure, there is the risk of respondents giving ill considered answers or they may even refuse to answer. 3e can reduce the number of comparisons per respondent either by presenting to each one of them only a sample of stimuli or by choosing a few objects which cover the range of attractiveness at about e ual intervals and then comparing all other stimuli to these few standard objects. Thus, paired!comparison data may be treated in several ways. $f there is substantial consistency, we will find that if R is preferred to N, and N to b, then R :*

will consistently be preferred to b. $f this is true, we may take the total number of preferences among the comparisons as the score for that stimulus. $t should be remembered that paired comparison provides ordinal data, but the same may be converted into an interval scale by the method of the ;aw of @omparative Iudgement. This techni ue involves the conversion of fre uencies of preferences into a table of proportions which are then transformed into matrix by referring to the table of area under the normal curve. The method is known as the @omposite 1tandard -ethod and can be illustrated as under2 1uppose there are four proposals which some union bargaining committee is considering. The committee wants to know how the union membership ranks these proposals. <or this purpose a sample of #== members might express the views as shown in the following table2 T!-,& :.42; R&sp#$s& P!""& $s #. 10# s!2p,&s P!i &% C#2p! is#$s 4 B !$%s D %=U '% += ! #&% 4 =.'((= (!).## =.==

S+11&s"i#$ A B C 6 P close up ! *(U &% E P Cepsodent '= ! &: @ P 6 uafresh '( (= ! / P @olgate Total := %= D: T7T6; #*( #&( #*: U Read as %= samples preferred Erand / to Erand 6. R!$4 O %& -c bj Rj 2 =.(&+( =.=D =.%= 3 =.'*%( (!)=.D =.=% 1 =.('(= =.## =.%%

@omparing me total number or preferences for each or the four brands, we find that @ is the most popular, followed by 6, E and / respectively in popularity. The rank order shown in the above table explains all this. Ey following the composite standard method, we can develop an interval scale from the paired! comparison ordinal data given in the above table for which purpose we have to adopt the following steps in order2 (i) Msing the data in the above table, we work out the column mean with the help of the formula given below2 @ e .( (?) #*( e .( (#==) -c P !!!!!!!!!!!!!!! P !!!!!!!!!!!!!!!!!!!!! P (.&+( n? '(#==) :+

where -c P the mean proportion of the columns @P the total number of choices for a given suggestion n P number of stimuli (proposals in the given problem)"brands ? P number of items in the sample. The column means have been shown in the -c row in the above table. (ii) The b values for the -c are secured from the table giving the area under the normal curve. 3hen the -c value is less than .(, the b value is negative and for all -c values higher than .(, the b values are positive. These b values are shown in b j row in the above table. 6s the bj values represent an interval scale, )ero is an arbitrary value. 9ence we can eliminate negative scale values by giving the value of )ero to the lowest scale value (this being (!).## in our example which we shall take e ual to )ero) and then adding the absolute value of this lowest scale value to all other scale items. This scale has been shown in Rj row in the above table. Araphically we can show this interval scale that we have derived from the paired! comparison data using the composite standard method as follows2

(iii)

@.@

@.C <ig. +.'%

@.A

@.B

@.?

(b)

-ethod of rank order2 Mnder this method of comparative scaling, the respondents are asked to rank their choices. This method is easier and faster than the method of paired comparisons stated above. <or example, with #= items it takes '( pair comparisons to complete the task, whereas the method of rank order simply re uires ranking of #= items only. The problem of transitivity (such as 6 prefers to E, E to @, but @ prefers to 6) is also not there in case we adopt method of rank order. -oreover, a complete ranking at times is not needed in which case the respondents may be asked to rank only their first, say, four choices while the number of overall items involved may be more than four, say, it may be #( or %= or more. To secure a simple ranking of all items involved we simply total rank values received by each item. There are methods through which we can as well develop an interval scale of these data. Eut then there are limitations of this method. The first one is that data obtained through this method are ordinal data and hence rank ordering is an ordinal scale with all its limitations. Then there may be the problem of respondents becoming careless in assigning ranks particularly when there are many (usually more than #=) items. Di &)" R&sp#$s& A""i"+%& S)!,&s ::

:.43

6ttitude scales help a researcher to directly measure an individualKs attitude or a component of the attitude. $n this techni ue, a consumer is re uired to respond to the scale being used, by explicitly stating his or her attitude towards a particular brand or brands, <or example, if a consumer is asked to rate the taste of ?escafe after being given a number of alternatives like, very good, good, neither good nor bad, bad and very bad, then it is said that a consumer is given a scale to measure the attribute. These responses are then evaluated as direct reflections of attitudes or a few components of attitudes. These scaled responses are termed as Kdirect response attitude scalesK. /irect response attitude scale is divided into two categories based on the number of attitude components measured. The first category is called Krating scalesK where a single dimension of components of attitudes is considered. <or example, pleasantness of a taste. The second category is called attitude scales, where several or all aspects of an individualKs attitude towards a product (object) are measured. 1o, we can infer that attitude scale is a combination of rating scales. :.43/!0R!"i$1 S)!,&s $n this method, the respondent places the attitude of the product (object) that is being rated at some point along a numerical series. The emphasis of the rating scales is on a) 6n overall attitude towards an object, such as /iet @oke. b) The existence level of a particular attribute in an object, like sweetness. c) 6n individualKs feelings towards an attribute of an object, such as liking the taste, or d) The importance associated .with an attribute, as the absence of caffeine. 6s a brandKs reputation may influence an individualKs preference towards a particular product attribute, such as taste, color, etc. the measurement of such products is usually carried out by removing the brand name. These tests are thus termed as blind tests. :.: ILLUSTRATIONS

(#) @onstruct following scales of attitude measurement (i) ?ominal (ii) 7rdinal (iii) $nterval (iv) Ratio (v) Thurstone (i) ?ominal 1cale 2 1ymbols or numbers are assigned to brand names, geographic territory, sex, user status, etc. I,,+s" !"i#$ 2 (a) ?ominal scale to identify potential of cellular phone (3;;) w.r.t. territory <ollowing data is provided on 3;; -obile Telephone Msers Q city wise Ci"3 M#-i,& T&,&p(#$& Us& s S32-#, :D Ci"3 M#-i,& Us& s S32-#,

#. Cune %. -umbai &. ?asik '. /elhi (. ;ucknow *. @alcutta

:==== &===== (==== %(==== *==== %=====

3 3 3 ? ? F

+. Curi :. 9yderabad D. Eangalore #=. @hennai ##. @ochin #%. Cunjim

(==== #:==== %+==== %+(=== '==== (====

F 1 1 1 1 3

3e can put the respective city in respective region like west, east, south and north and put the first word of region against each city name. 3e now add potential users under 3,F,1 H ? and represent as follows2 N# "( T& i"# 3 &#==== A&s" T& i"# 3 ':==== S#+"( T& i"# 3 +%(=== E!s" T& i"# 3 %(====

C#$),+si#$ ;9 6ttitude formed is, south territory has highest potential. I,,+s" !"i#$ (b) Aodrej 6grovet have provided following data for its 4cattle fee5 product Erand 4-ilk -ore5. @onstruct ?ominal scale. Dis" i)" #. ?agpur %. 6kola &. 3ardha '. 6urangabad (. Ialna *. Carbhani R&1i#$ Bidharbha /o /o -arathwada /o /o S!,&s p& %!3 i$ C",. &== %== '== #(= #=( #%( Dis" i)" +. 1olapur :. Earamati D. 1angli #=. Cune ##. ?asik #%. 1atara R&1i#$ 1outh -aharashtra /o /o 3et -aharashtra /o /o S!,&s p& %!3 i$ C",. %*= &'= %:= (== &:= '==

;et us regroup regionwise sales and rank them S . N#. # % & ' R&1i#$! Bidharbha -arathwada 1outh -aharashtra 3est -aharashtra T#"!, s!,& i$ C",s. P& %!3 D(= &:= ::= #&&= R!$4 $$ $B $$$ $

C#$),+si#$ 2 6ttitude formed is 3estern -aharashtra Region is having highest sales potential.

D=

(ii) # %i$!, S)!,& 2 $n marketing research, ordinal scales are used to ascertain consumer5s perception on a brand, service, etc. $llustration a) -obile user5s brand preference for handset manufacturers 6ttributes ;igh t weig ht Crice /esign 9igh " style technolo gy Eatter y life /urabi lity Reliabili ty Boice ualit y @aring Total compa score ny in > DD> DD> D+> D&> :%> :=> *:> **> &=> #&> (>

?okia 1ony Fricson -otorola ;A 1amsung Canasoni c Chilips 1iemens -itsubhis hi 6lcatel T@; I,,+s" !"i#$ (b)2 -icrowave ovens manufactures wants to know the brand ranking perceived by customers. /esign ordinal scale. 6ttributes ;A EC; Genstar 1amsung Flectrolux Canasonic ?ational 3hirlpool Eajaj Crice(#=) D : #= : * * + + : 3eight (#=) D : #= : D : + + + 6ntibacterial properties (#=) #= : + #= D : + + #= ?utritive food (#=) #= : : #= : : + + #= 6fter sales service (#=) #= : #= #= * * + + #= Total service out of (= ': '= '( '* &: &* &( &( '(

Rank # ' & % ( * + + &

(iii) I$"& '!, s)!,&2 $n marketing research, this scale is used to measure intensity by which attitude towards a brand varies o any marketing stimuli. I,,+s" !"i#$2 (a) -obile telephone users may express ?okia brand in follows2 Alobal brand ?okia cellular $ liked by me the most, $ neither like nor dislike ?okia cellular, $ dislike ?okia cellular, $ dislike ?okia cellular the most. D#

$llustration (b) @onsumers want to express the tastes (likes " dislikes) of Ci))as and burgers supplied by Ci))a 9ut, -c/onalds and domino Ci))a. /esign $nterval 1cale. A"" i-+"&s Taste @heese 0uality Eriskness Thickness 1pice Crice C#$),+si#$2 -c/onald5s Eurgers is liked by most of the consumers. (6mul Ci))a is not considered because only in Aujarat, it is served in ready to eat fashion whereas in other part of the country it is served in fro)en condition, which re uires further processing). (iv) R!"i# s)!,& 2 This scale is used to measure attitude on consumers, profitability, probability of purchase, etc. uantity sold, number of Li4& "(& 2#s" P D M Li4& P D M N&i"(& ,i4& $# Dis,i4& P D M Dis,i4& P D M Dis,i4& "(& 2#s" P D M

I,,+s" !"i#$ 2 (a) $T customers handled by the Telecom companies. Q ?umber of $T consumers handled by Tata $ndicom is one tenth of that handled by Reliance $nfocom. $llustration (b) 6 automobile dealer wants to get knowledge on profitability on consumer base of hero 9onda and TB1 Bictor. /esign ratio scale2 V&(i),& 9ero 9onda Cassion TB1 Bictor A; C+!$"i"3 s#,% i$ #$& 3&! %&= %=( P i)& ?&!)( i$ Rs. '*=== '(=== T#"!, s!,& Rs. #=(:==== D(%(=== C#22issi#$ &! $&% p& '& "i)!, '=== '(== T#"!, p #.i" Rs. D%==== D%%(==

C#$),+si#$s2 (#) ?o. of consumers handled by 9ero 9onda /ealer are #.#% times more than TB1 victor dealer (%) Total profitability of TB1 Bictor dealer is #.==%+ times more than 9ero 9onda /ealer. (v) Thurstone scale 2 This is eleven point scale to express varying degree of attitude from unfavourable to favourable. E 6 < G @ / F A 9 $ I Mnfavourable ?eutral <avourable $llustration (a) suppose a statement is made like, 4Fnron Cower Croject is beneficial to $ndia5, the response from the consumers, politicians and govt. could vary from #==> unfavourable to #==> favourable. D%

$llustration (b) /esign Thurstone scale for 41aas!Eahu5 TB serials being run on most of the prime channels (1@9-9R/ -ay %==(() S#,+"i#$ 2 <ollowing statements (from 6 to G) could be made. (a) (b) (c) (d) (e) (f) (g) (h) (i) (j) (k) (%) 6ll these 1aas!Eahu serials build up negative value system by depicting disputes in the family. 6ll 1aas!Eahu serials portray an irrational depiction of characters. Telecast time of 1aa!Eahu serials clashes with important programmes like ?ews, etc. 6ll 1aas!Eahu serials are monotonous. -ost of the 1aas!Eahu serials are complete waste of time. $ have no positive or negative feelings about 1aas!Eahu serials, 1aas!Eahu serials provide good entertainment after a hard days work. -ost of the key characters of 1aas Eahu serials become trend setters in respect to clothings, jewelry and other accessories. -ost of the 1aas Eahu serials bring the whole family together 1aas!Eahu serials help to understand, analyse and solve the domestic crisis. The 1aas!Eahu serials are a good ways to instill family values (obedience respect) @ompare rating and ranking scales (1@9-9R/ /ec. %==()

S#,+"i#$s2 R!"i$1 S)!,& R!$4i$1 S)!,& #. 6ttitude is measured from the point of #. 6ttitude is measured from the point of view of view of intensity of the likes and dislikes intensity of preferring one product over other. %. $nterval data is needed %. 7rdinal data is needed &. $t is absolute &. $t is relative '. Fxamples Q $nterval 1cale '. Fxample Q 7rdinal scale, 1emantic differential scale (&) (i) @onstruct ;ikert 1cale, Cerceptual -ap and 1emantic /ifferential 1cale ;ikert 1cale to study consumer satisfaction with tyre!brands e% 1trongly 6gree e# 6gree = !# ?either agree /isagree nor /isagree !% 1trongly disagree

1core out of 6ttributes @ost friendly Arip 3ear H tear Rubber Q uality

D&

longevity B !$% -R< IG 6pollo Si1$ S)# & %e%e%e%e% P #= %e#e%e#e%P: %e#e=e=e=P&

(ii) 1emantic differential scale to understand the images in the mind of consumers for washing machine manufacturers Remark Fxcellent Eetter e% B Aood e# ;A 6verage Coor = !# 6ttributes Traditional Mnreliable 3eak ?on cust! focused ?on response -ore poor 3orst

1core e& 6ttributes 3 Crogressive Reliable 1trong @ust! focused Responsive Erand 1ign 3hirlpool 3 Bideoon B

1core e&e%e%e&e% P #& e%e%e&e%e%P##

;A ;A e&e%e%e%e%P## (iii) <ollowing data is given for three Telecom companies. Crepare 1emantic /ifferential 1cale 6ttributes R$E96R6T$ T6T6 1olution 2 e& Reliability Tangibility Responsiveness 6ssurance Fmpathy D' e% e# = !# !% !& Reliability e& e& !% Tangibility e% e% !& Responsiveness e# e% = 6ssurance e# e& e# Fmpathy e# e& e#

B !$% R$E6

Si1$

S)# & e&e%e#e#e% P : e&e%e%e&e&P#& !%!&e=e#e#P!&

Eharati E9 Tata (iv) T6 Cerceptional -ap

<ollowing data is given on $ndian Refrigerator $ndustry (1i)e #*( lit. to #:= lit.) Erand EC; Aodrej Gelvinator 1amsung 3hirlpool ;A Flectrolux Bideocon 6llwyn Boltas /aewoo Crepare Cerceptual map. Technology /irect cool /o /o /o /o /o <rost free /irect cool /o /o <rost free Crice 2Rs. D%D= :=== DDD= :'D= D#== D=== ##=== ::D= :%D= :##= #=(==

S#,+"i#$2 @onstruction of Cerceptual -ap

&igh price (5s.) CA@@@ CCE@@ CC@@@ C@E@@ /irect cool 9hirlpool Qelvinator >43 31 DE@@ D@@@ 0rost "ree

Samsung

D(

Alwyn

IE@@ 2oltas 1ordre6 I@@@

3ow 4rice

D*

CHAPTER 7 OBSERVATION RESEARCH 7.: HOA OBSERVATION RESEARCH IS USEDH (i) (ii) (iii) 7.:1 (i) (ii) (iii) (iv) (v) (vi) The major factor that influence the sale of branded goods is its availability. 9ence marketer may depute observer to some of stores to find out how fre uently the product is out of stock. -arketer can also observe the prices charged by competitors. -arketer can use people5s meter to understand reaction of consumers to brand features.

M&"(#%s #. #-s& '!"i#$ 1tructured Q $t is used when the research problem is already formulated precisely and the observers have been told specifically what is to be observed. Mnstructured Q 7bservers are free to observe whatever they think is relevant and important /isguised Q 1amples do not know that they are being observed. Mndisguised Q 1amples are informed about motives of observers /irect Q Eehaviour of sample is observed as it occurs $ndirect Q 1ome record of past behaviour is observed. $n other words, the behaviour itself is not observed rather its effects are observed. A%'!$"!1&s #. #-s& '!"i#$ R&s&! )(

7.:2

(i) 1ince samples are observed directly, there is no chance of bias on account of predisposition of any kind (ii) The willingness of sample is not essential, as such lot and time is saved in seeking permissions (iii) Rural samples are difficult to top due to communication problem, for whom this method is not suited. 7.:3 Dis!%'!$"!1&s

(i) The action being observed of the samples may not provide any reasons for his behaviour, attitude and opinion (ii) The particular action observed might not be in actual circumstances but may be only on one particular day. (iii) The data collected dependent on skill of the observer and the manner in which he needs and interprets it. 1ame results might not hold with different set of observers. 9ence this method is not very reliable. (iv) Bery few samples can be observed due to time constraint

D+

CHAPTER = SCHEDULE AND CUESTIONNAIRE S!2p,i$1 M&"(#%s <or probability sampling techni ue (i) (ii) (iii) (iv) (v) Random 1ampling 1ystematic 1ampling 1tratified 1ampling -ultistage 1ampling 6rea 1ampling

<or non!probability techni ue (i) (ii) Curposive 1ampling 0uota 1ampling

C+&s"i#$$!i & D&si1$ !$% D !."i$1 <ollowing elements in balance manner make good uestionnaire (i) (ii) (iii) (iv) The uestions must be relevant to subject matter and a set of uestions must be able to cover the entire topic of the research (illustration of @haitanya 9ealth @lubs uestionnaire) The uestion should not indicate specific answers. (example of 6mul5s -asti curd and 9;;5s study on 1urf 3ash Eoosters) ;engthy and difficult uestions would lose customer attention and hence short and easy uestions to be posed. Fach and every uestion should create interest in the minds of samples so that samples also feel importance of uestion being asked and hence likely to give accurate answers seriously.

D:

App #p i!"& L!3#+"

.n"ormation -eeded (Secondary or 4rimary data)

*ethod o" data collection, 4., T. or +bservation

Sampling technique and methods

Muestionnaire 3ayout /esign /ecide on content o" each question /ecide on type o" questions /ecide on wording or questions /ecide sequence o" questions /ecide pre:testing o" questionnaire

0inal testing o" revised questionnaire

T3p&s #. C+&s"i#$s (#) (%) (&) 7pen!ended uestion 3hat do you think of the test of brand R cola8 /ichotomous uestions (a) 6re you user of R toilet soap8 Nes " ?o. (sometimes one more answer is provided like ,any other please specify.) -ultiple @hoice uestions 3hich of the following factors made you by this brand of car2 (a) Reasonable price (b) Areat looks (appearance) (c) <uel economy (d) Fasy availability of service (e) 6ny other, please specify. Ratings or Rankings Rating uestions (a) Clease rate the following detergent rank on a scale of #! + in their ability to clean clothes Erand 6 Erand E # # % % & & DD ' ' ( ( * * + +

(')

(()

Erand R # % & ' ( * + (b) Clease rank for following detergents on their ability to clean the clothes Erand 6 # % & ' ( * + Erand E # % & ' ( * + Erand R # % & ' ( * + (# means best, % means better, & means good, !!!!!!!, + means worse) $ndirect uestions (a) -ost of the people in $ndia smoke N#$9Fi,"& Ci1! &""&s because !!!!!!!!!!!!!!!! (b) Io Eibi 1e kare pyar wo !!!!!!!! se kaise kare inkar8

$llustration2 @onstruct a uestionnaire for understanding buyer behaviour in 1election of television set for household segment 7bjectives2 (i) 3hat features buyers are looking for in a TB set (ii) 9ow important the price to the buyer (iii) 3hat are the methods of payment (iv) The selection process of the buyer 0uestionnaire2 (#) a. /o you own a television8 Nes " ?o b. $f yes, which brand " company name c. $f no, go to uestion +. (%) 3hile buying a TB what are the features you look for. (&) Aiven below some of the features of the TB. 9ow important is each one to you, please tick mark. <eatures Fxtremely $mportant 1ome what ?ot very ?ot important important important important at all (i) ;ooks (ii) Cortability (iii) @abinet ! -oulded ! 3ooden (iv) 1i)e of the screen (v) ?o. of channels (vi) ?o. of speakers (vii) 6uto control monitor (viii) -anufactures reputation (ix) Bideo adaptability (x) $ntegral /B/ #==

(xi) <oreign collaboration (xii) Auarantee offered (xiii) 1ervicing arrangement (xiv) Crice (xv) @hild lock (xvi) Aames (') a. $f a price of TB is classified as high, medium and low then where your TB model belongs to8 b. 9ow do you judge price of a TB with respect to the features of a TB8 c. 3hich payment option do you prefer8 @ash " $nstallment 6t the time making brand choice decision, from whom among other following sources did you take the advice8 <amily members <riends " neighbours /ealers 6dvertisement 6ny other source, please specify 6 set of statements are given below. Clease indicate your opinion, to be recorded a scale ranging from strongly agree to strongly disagree 1trongly 6gree 6gree @an5t say /is!agree 1trongly disagree

(()

(*)

1tatements (i) Cossessing TB set is a status symbol (ii) 7bserving TB is passing time (iii) //s TB programmes are dull whereas @ H 15s programmes are attractive (iv) TB affects children education (v) $ndian TB programmes are educative (vi) TB is best source of entertainment (vii) TB is low cost entertainment (viii) Aovernment5s decision on expanding TB network through /T9 and dish TB is appreciable (ix) TB is best gift item (x) people are confined to dairy homes due to TB viewing

#=#

(+)

@lassify data (a) 6ge (b) Fducation (d) 6nnual income of the family

(c)

7ccupation

#=%

CHAPTER 10 INTERVIEA 10.0 INTRODUCTION

$n this chapter we will discuss how primary data is collected. Crimary data can be collected either by observation or by communicating with people. 1urvey research is one of the two alternatives for gathering primary data. $t is the process of systematic gathering of information from people for the purpose of understanding various aspects of the population being studied. $t involves uestioning the concerned people (a sample) and recording their responses. These responses are further analy)ed to arrive at a conclusion. Fxcept in a few instances, this method has proved to be the most efficient way of collecting data. $nformation regarding the attitudes and behavior of consumers can be best collected through surveys. The main issues concerning a survey research are sampling, uestionnaire design, and uestionnaire administration and data analysis. 9owever, this method has some drawbacks. Through this chapter we attempt to make a comprehensive study of different types of survey research, its merits and demerits. 10.1 TYPES OF INTERVIEAS

The interview is one of the most important components of marketing research. $nterviews can be classified into structured and unstructured and direct and indirect. 10.2 TYPES OF SURVEYS OR PRIMARY DATA COLLECTION METHODS

There are mainly four types of surveys2 Cersonal, Telephone, -ail and @omputer. 10.21 P& s#$!, I$"& 'i&5s ? M!i, M&"(#% ? T&,&p(#$& I$"& 'i&5 M&"(#% M&"(#%s #. %!"! )#,,&)"i#$ i$ .i&,% &s&! )( Flement of differentiation ?o of samples Cersonal method interview -ail method Telephone interview method

?ot very high due time constraint

;arge no samples can be contacted

-uch more sample can be contacted as less time re uired

Time

$s used when ade uate time is available 9ighest

Msed when considerable time is available ;owest

Msed when very short is available -oderately high as compared to --

@ost

#=&

6ccuracy

9ighest due to personal interaction and data recording with right understanding

Mse

$nfrastructure

?ot much useful when large geographic area is to be cover due to cost constraint 9uge infrastructure in form of project leader, research officer and investigators re uired

?ot very high due to a. Response rate not more than %= to %(> b. 3rong interpretation of 0us. can not be sorted out. <or geographically scattered samples this is best suited 6lmost negligible

<airly high but depends on skill of interviewer in sorting out misinterpretation of 0us.

<or outstation samples the cost could be prohibitive hence useful local surveys only. $n terms of skilled telephone operator and data base

Type of samples

Mseful for ignorant and illiterate samples 1amples loose interest with lengthy uestionnaire

1uitable for samples who can read and write lengthy uestionnaire is no prob. bco) 1ample feel it at his convenient time

1uitable if samples properly communicate

can

0uestionnaire

;egthy uestionnare wont do bco) sample is not directly seen

$nterviewer

1killed $nterviewer can 1killed o not skilled improve accuracy $nterviewer r does not affect accuracy

1killed $nterviewer can improve accuracy

Type of 0us

1uitable for 1uitable for 1uitable only for 1pontaneous 6ns spontaneous 6ns since spontaneous as well as samples do not like to well thought 6ns tax their memories

#='

/rawbacks";imitations $f investigators are not 0uestionnaire might not $t is impossible to judge person trend, he himself might be filled up by intended contacted is desired person as such fill up 0uestionnaire person the #st name of sample must be known

$nternet $nterviewing 3eb interviewing Fmail survey f 6pplications f 6ll type of Fxam marketing surveys f @6T f 6dmissions A%'!$"!1&s #. I$"& $&" i$"& 'i&5i$1 f f f f f f f f f f <ast set up, Fxecution and completion Bisual stimuli can be evaluated ( in case of web cap) 1timuli presentation can be controlled allowing for pre and post uestions unlike traditional mail. 0uestion presentation is consistent and eliminates interviewerKs bias 0uestionnaire skip pattern can be controlled ;ess instructive process, allowing respondents to ans as per their convenience 6ccurate responses possible since it is self administered. Fliminates cost of an interviewer Cermits real time data -uch cheaper than traditional research procss.

S!2p,i$1 f a. b. c. d. e. f. 3eb interviewing can be generated provided sample are accessible. 9ence samples to b chosen from Bisitors to a website F!commerce customers Msers of certain compuer hardware or software Fmployees of a company that provides web access for %' hours Regular web surfers at net cafes 3ap users

#=(

S+ '&3 %&si1$ )(! !)"& is"i)s f f f f f f 1creen look and feel 0uestion layout 3ord dynamics for onscreen uestionnaires5 Clacement of graphics Randomi)ation of ans or stimuli Richer open end responses

App,i)!"i#$s 1tudy of competitorKs product where C$"--"T$ may not be possible due to cost, time Fct. f -arketer can heir a .com company to study competitorKs actions Li2i"!"i#$s f f f ?o. of C@ owner"internet users are limited This type of survey can not be for masses but for classes 1ampling is complex due to problems in identifying in sample frame. f

#=*

CHAPTER 12 PROCESSING AND ANALYSIS OF DATA D!"! A$!,3sis !$% P #)&ssi$1 #) /ata Crocessing 2 $t involves editing, coding, classification and tabulation of collected raw data so as to facilitate analysis. %) /ata 6nalysis2 To identify the relationships in the processed data to statistical tests of significance in order to establish the validity of these relationships, either to support or reject research hypothesis. /ata Crocessing #) Fditing 2 $t is a process of examining the collected raw data to detect errors and omissions and to correct these when possible. $t involves a careful scrutiny of completed uestionnaire. $t is also done to assure that the data is accurate, consistent, complete and homogeneous a) <ield Fdition Q @orrecting the abbreviations or illegible uestionnaire re uired from field at researcher5s office. b) @entral Fdition Q Fditing uestionnaires which are filled at investigator5s office by the investigators. %) @oding 2 $t is the process of assigning numbers or other symbols to answers so as to transform the ualitative data into uantitative terms. Fxample 0uestion 3hat is your material status8 Cossible 6nswer ?o 6nswer -arried Mn!-arried 3idow " 3idower /ivorcee @odes # % & ' (

12.4

CLASSIFICATION

$t is the process and arranging the data according to points of similarities and dissimilarities2 E<!2p,& (#) Aeneral Fducation 2 (a) @lassification as per attributes #=+

 (%) U

;ess than 91@ 9igh school 91@ completed Araduate @ollege Cost Araduate /octorate Crofessional Fducation /iscipline 1ocial 1cience (-13) -edicine <inance Technology " Fngg. Eachelor5s level -aster5s level /octorate level

U (b) :

Croficiency

@lassification as per class!intervals /ata relating to income, production, age, weight, etc. $ncome 1F@ 6 -onthly income Rs. %=,===e 1F@ E -onthly income Rs. #(===e 1F@ @ -onthly income Rs. #====e

6ge #=!%= Q #= and under %= %=!&= Q %= and under &= &=!'= Q &= and under '= 12.6

TABULATION AND TYPES AND TABLES

/efinition 2 6 statistical table is the logical listing and uantitative data in various rows and column with self explanatory title, row headings column headings and notes regarding source of data, context, etc. 7bjectives 2 1implifying data, avoid unnecessary data, facilitate comparison 12.61 T3p&s #. "!-,&s 12.611 O$& 5!3 T!-,& I Si$1,& )(! !)"& is"i)s

T!-,& N#. 2 9eading 2 @lasswise distribution of ?o. of students in $1-1 and $$#=:

@lass $1-1 (a) CA (b) -asscom $$(a) (b) (c) (d) -E6 -@6 -@E@6

?o. of students *== #%= +%= #D: *= D= #== '': ##*:

@olumn 9eading

Row 9eading

Arand Total 1ource2 Registrar " $1-1 " Near %==( 12.612

T5#95!3 T!-,& I T5# )(! !)"& is"i)s

Table & 9eading2 @lass wise and sex wise distribution of no. of students in $1-1 and $$S&< C,!ss PG (== MBA #D= 1ource2 Registrar $1-1 (%==() 12.613 T( && 5!3 "!-,& ; T( && )(! !)"& is"i)s 9eading 2 @lass wise, <aculty wise and sex wise distribution of no. of students in $ndira @ollege @lass PG F M&%i)!, M F F!)+,"3 C#2p+"& F #== =: *== #D: M!,& F&2!,& T#"!,

M T T M T #st year %nd year &rd year 'th year <oot note2 -!-ale, <!<emale, T!Total 1ource 2 Registrar Q $1-1 " year %==( 12.@ GRAPHICAL REPRESENTATION OF STATISTICAL DATA ; HISTOGRAM

M!ss C#22. M F T

#=D

$t represents fre uency of distribution in series and adjacent rectangles erected on R axis with class interval as base, hence width of rectangle is e ual to class width. 6rea of rectangle is taken as proportional to class fre uency. $llustration 2 Aive below data on land holding by farmers in 9aveli Taluka 1i)e of farm hectares ?o. of farms in #!%= #% %#!'= &: '#!*= #* *#!:= ( :#!#== & #=#!#%= #

?@ BE B@ fre7uency AE A@ CE C@ E @ A@

1cale R axis 2 # unit P %= hect. N axis 2 #unit P ( farms

?@

H@

I@

C@@

Si6e of farms in hect

$llustration 2 Represent following data by 9istogram 6ge Aroup (in yrs.) Copulation =!( (== (!%= %#== %=!&= %%== &=!'= %=== '=!*= #*==

*= '

1ince classes are of une ual width, convert the data by computing fre uency density. 6ge Aroup (in yrs.) Copulation =!( (=="(P#== (!%= %#=="#( P #'= %=!&= %%=="#= P %%=

##=

AE@ fre7uency of distribution A@@ CE@ C@@ E@ @ 1cale R axis 2 # unit P ( yrs. N axis 2 #unit P ( units of fre uency density

C@

A@

B@

?@

E@

H@

#@

I@ D@

C@@

12.:

POLYGON (fre uency polygon)2

$n this representation, mid values of fre uency " data are taken on R axis and fre uency o the N axis -onthly 9ouse Rent ?o. of families -id values will be <re uency will be
B@ AE #o. of fami*ies A@ CE C@ E @ @ A@@ ?@@ H@@ I@@ C@@@ CA@@ C?@@ CH@@

#==!&== * %== * '== #*

&==!(== #* *== %' :== %=

(==!+== %' #=== #= #%== '

+==!D== %= (R axis) (N axis)

D==!##== #=

##==!#&== '

12.7

BAR DIAGRAM

3idth of bar indicate specific attribute in case of one!dimensional Ear diagram ###

Aiven below dividend declared by $nfosys construct bar chart Near =(!=* ='!=( /ividend > #%==> #%==>
)ividend 8 by !nforsys
C?@@K CA@@K C@@@K I@@K H@@K ?@@K A@@K @K @C:@A @A:@B 4ears @B:@? @?:@E #E@K I@@K C@@@K

=&!=' #===>

=%!=& :==>

CA@@K

1ub divided bar diagram

$250 bn $200bn $150bn $100bn $50bn I $130 bn

Cost and profit

I $100 bn

RM 50% OH 20% 30% MARGI N

E -ultiple bars 2 (a) 1ales Turn!over Bs Targeted 1ales for ;A $92bn $80 bn E

A@@E:@H

CH@@ D#@@ B@@@ B@@@ EE@@ E@@@ I@@@ #@@@ @ A@@@ ?@@@ H@@@ I@@@ C@@@@ CA@@@

A@@?:@E

A@@B:@?

(b)

A@@A:@B

1ales, AC, ?C

##%

CH@ C?@ CA@ C@@ I@ H@ ?@ A@ @ A@@? ?@ A@ CA@

CBE

BE CE

A@@E

#umber of cities covered by different air*ines

#o. of cities covered E@ ?@ B@ A@ C@ @ spice6et air deccan +ir*ines 6et airway

12.=

PIE DIAGRAM

Aiven below market share for @T construct pie diagram ;A 7nida Bidecon 1amsung 7thers The angle at centre is given by Q !!!!!!!!! x &.* #==
Samsung, CHK

%*> %%> %=> #*> #*>

%* x &.* %% x &.* %= x &.* #* x &.* #* x &.*

D&.* +D.% +% (+.* (+.*

+thers, CHK 31, AHK

Cercentage !!!!!!! x &*= P percentage

##&
+nida, AAK 2idecon, A@K

12.=01 S"!"is"i)!, M!ps ; These are nothing but the histograms, etc. generated by using 1C11 software. 12.=1 I,,+s" !"i#$s Araphs of fre uency distributions

figures like pie chart, bar charts,

6 fre uency distribution can be presented graphically in any of the following ways2 9istogram <re uency polygon 1moothed fre uency curve and 7gives or cumulative fre uency curves His"#1 !2 7ut of several methods of presenting a fre uency distribution graphically, histogram is the most popular and widely used in practice. 6 histogram is a set of vertical bars whose areas are proportional to the fre uencies represented. 3hile constructing histogram the variable is always taken on the R!axis and title R!axis and the fre uencies depending on it on the N!axis. Fach class is then represented by a distance on the scale that is proportional to its class!interval. The distance for each rectangle on the R!axis shall remain the same in case the class!intervals are uniform throughout. $f they are different they vary. The Y-axis represents the fre uencies of each class which constitute the height of its rectangle. $n this manner we get a series of rectangles each having a class!interval distance as its width and the fre uency distance as its height. The area of the histogram represents the total fre uency as distributed throughout the classes. The histogram should be clearly distinguished from a bar diagram. The distinction lies in the fact that where a bar diagram is one dimensional. i.e. only the length of the bar is material and not the width. 6 histogram is two!dimensional that is in a histogram both the length as well as the width are important. The histogram is most widely used for graphical presentation of a fre uency distribution. 9owever we cannot construct a histogram for distribution with open!end classes. -oreover a histogram can be uite misleading if the distribution has une ual class!intervals and suitable adjustments in fre uencies are not made. The techni ue of contracting histogram is given below (i) for distributions have e ual class! intervals and (ii) for distributions having une ual class!intervals. 3hen class!intervals are e ual, take fre uency on the Y-axis, the variable on the R!axis and construct adjacent rectangles. $n such a case the height of the rectangles will be proportional to the fre uencies.

##'

3hen class!intervals are une ual, a correction for une ual class intervals must be made. The correction consists of finding for each class the fre uency density or the relative fre uency density. The fre uency density is the fre uency for that class divided by the width of that class. 6 histogram or fre uency density polygon constructed from these density values would !have the same general appearance as the corresponding graphical display developed from e ual class intervals. <or making the adjustment we take that class which has lowest class!interval and adjust the fre uencies of other classes in the following manner. $f one class!interval is twice as wide as the one having lowest class!interval we divide the height of its rectangle by two, if it is three times more we divide the height of its rectangle by three etc. i.e. the heights will be proportional to the ratio of the fre uencies of the width of the class. The following illustration would clearly indicate the manner in which the adjustment has to be made2 I,,+s" !"i#$s 2 Represent the following data by a histogram2 M! 4s =!#= #=!%= %=!&= &=!'= '=!(= N#. #. s"+%&$"s : #% %% &( '= M! 4s (=!*= *=!+= +=!:= :=!D= D=!#== N#. #. s"+%&$"s *= (% '= &= (

S#,+"i#$2 1ince class!intervals are e ual throughout no adjustment in fre uencies is re uired.
H!STO"%+(
#@ H@ E@ (arks ?@ B@ A@ C@ @
@:C@ C@:A@ A@:B@ B@:?@ ?@:E@ E@:H@ H@:#@ #@:I@ I@:D@ D@:C@@

H@ EA BE AA I CA E ?@ ?@ B@

#o. of students

I,,+s" !"i#$2 Represent the following data by means of a histogram (1$E- -ay %==() A&&4,3 5!1&s /i$ Rs.0 #=!#( #(!%= %=!%( N#. #. 5# 4& s + #D %+ ##( A&&4,3 5!1&s /i$ Rs.0 &=!'= '=!*= *=!:= N#. #. 5# 4& s #% #% :

%(!&=

#(

S#,+"i#$; 1ince the class!intervals, are une ual, fre uencies must be adjusted otherwise the histogram would give a misleading picture. The adjustment is done as follows. The lowest class interval is (. The fre uencies of the clas&=!'= shall be divided by two since the class interval is double, that of '=!*= by ', etc. @onstruction of 9istogram when only mid!points are given. 3hen only mid!points are given, ascertain the upper and lower limits of the various class and then construct the histogram in the same manner.
H!STO"%+(
B@ AE A@ CE C@ E @
C@:CE CE:A@ A@:AE AE:B@ B@:?@ ?@:H@ H@:I@

A# CD CE CA # CA I

0eek*y :a es ;in %s.<

I,,+s" !"i#$; /raw the histogram for the following data 2 (1@-9R/ -ay %==() V! i!-,& #==!##= ##=!#%= #%=!#&= #&=!#'= F &*+&$)3 ## %: &* 'D
H!STO"%+(
H@ E@ fre7uency ?@ B@ A@ C@ @
C@@:CC@ CC@:CA@ CA@:CB@ CB@:C?@ C?@:CE@ CE@:CH@ CH@:C#@

V! i!-,& #'=!#(= #(=!#*= #*=!#+=

F &*+&$)3 && %= :

?D BH AI CC BB A@ I

9ariab*e

F &*+&$)3 P#,31#$ ##*

6 fre uency polygon is a graph of fre uency distribution. $t has more than four sides. $t is particularly effective in comparing two or more fre uency distributions. There are two ways in which a fre uency polygon may be constructed2 3e may draw a histogram of the given data and then join by .straight lines the mid!points of the upper hori)ontal side of each rectangle with the adjacent ones. The figure so formed is called fre uency polygon. $t is an accepted practice to close the polygon at both ends of the distribution by extending them to the base line. 3hen this is done two hypothetical classes at each end would have to be included!each with a fre uency of )ero. This extension is made with the object of making the area under polygon e ual to the area under the corresponding histogram. The readers are advised to follow this practice. 6nother method of constructing fre uency polygon is to take the mid!points of the various class!intervals and then plot the fre uency corresponding to each point and to join all these points by straight lines. The figures obtained would exactly be the same as obtained by method ?o.#. The only difference is that here we have not to construct a histogram. Ey constructing a fre uency polygon the value of mode can be easily ascertained. $f from the apex of the polygon a perpendicular is K drawn on the R!axis, we get the value of mode. -oreover fre uency polygons facilitate comparison of two or more fre uency distributions on the same graph.

F &*+&$)3 p#,31#$ (!s )& "!i$ !%'!$"!1&s #'& "(& (is"#1 !2; The fre uency polygons of several distributions may be plotted on the same axis, thereby making certain comparisons possible, whereas histograms cannot be usually employed in the same way. To compare histograms we must have a separate graph for each distribution. Eecause of this limitation for purposes of making a graphic comparison of fre uency distributions, fre uency polygons are preferred. The fre uency polygon is simpler than its histogram counterpart. . $t sketches an outline of the data pattern more clearly. The polygon becomes increasingly smooth and curve!like as we increase the number of classes and the number of observations.

$n the construction of fre uency polygon the same difficulties are faced as with histograms i.e. they cannot be used for distributions having open!end classes and suitable adjustment as in case of histogram, it is necessary when there are une ual class!intervals. (E.@om., Cune univ., %==*) M! 4s =!#= #=!%= N#. #. s"+%&$"s ' * ##+ M! 4s (=!*= *=!+= N#. #. s"+%&$"s #' :

%=!'= '=!(=

#' #*

+=!D= D=!#==

#* (

H!STO"%+(
A@ #o. of students CE C@ E @ @:C@ C@:A@ A@:?@ ?@:E@ E@:H@ H@:#@ #@:D@ D@:C@@ (arks ? H C? CH CH

C? I

S2##"(&% F &*+&$)3 C+ '& 6 smoothed fre uency curve can be drawn through the various points of the polygon. The curve is drawn freehand in such a manner that the area included under the curve is approximately the same as that of the polygon. The object of drawing a smoothed fre uency curve is to eliminate as far as possible accidental variations that might be present in the data. 3hile smoothing a fre uency polygon the fact that it is really derived from the histogram should always be kept in mind. This would imply that the top of the curve would overtop the highest point of the polygon particularly when the magnitude of class!interval is large. The curve should look as regular as possible and sudden turns should be avoided. The extent of smoothing would however depend upon the nature of the data. $f it is a natural phenomenon like tossing of coin, smoothing may be freely resorted to as such phenomenon normally has symmetrical curves. Eut if the phenomenon is social or economic the curve is generally skewed and as such smoothing cannot be carried, too far. <or drawing a smoothed fre uency curve it is necessary to first draw the polygon and then smooth it out. 6s discussed earlier, the polygon can be constructed even without first constructing histogram by plotting the fre uencies at the mid!points of class!intervals. This may save some time but the smoothing of the polygon cannot be done properly without a histogram. 9ence it is desirable to proceed in a se uence i.e. first to draw a histogram then a ##:

polygon and lastly to smooth it to obtain the smoothed fre uency curve. This curve should begin and end at the base line and as a general rule it may be extended to the mid!points of the class!intervals just outside the histogram. The area under the curve should represent the total number of fre uencies in the entire distribution.

The following points should be kept in mind while smoothing a fre uency curve2 7nly fre uency distributions based on samples should be smoothed. 7nly continuous series should be smoothed. The total area under the curve should be e ual to the area under the original histogram or polygon.

$llustration 2 Represent the following fre uency distribution by means of a 9istogram d superimpose thereon the corresponding fre uency polygon and fre uency curve2 1alary (Rs.) &==!'== '==!(== (==!*== *==!+==
C?@ #o. of 'mp*oyees CA@ C@@ I@ H@ ?@ A@ @
B@@:?@@ ?@@:E@@ E@@:H@@ H@@:#@@ #@@:I@@ I@@:D@@ D@@:C@@@ C@@@: CA@@

?o. of employees %= &= *= +(

1alary (Rs.) +==!:== :==!D== D==!#=== #===!#%==

9istogram

?o. of employees ##( #== *= '=

<re uency curve <re uency Colygon

sa*ary ;%s.<

C+2+,!"i'& F &*+&$)3 C+ '&s # O1i'&s 6t times we are interested in knowing how many workers of a factory earn less than Rs. +== per month or Khow many workers earn more than Rs. #,=== per monthK or Kpercentage of students who have failedK, etc. To answer these uestions, it is necessary to add the fre uencies. 3hen, fre uencies are added, they are called cumulative fre uencies. These fre uencies are then listed in a table called a cumulative fre uency table. The curve obtained ##D

by plotting cumulative fre uencies is called a cumulative fre uency curve of an 7give (pronounced 7jive). There are two methods of constructing 7give, namely2 (a) (b) (a) (b) The Kless thanK method, and the Kmore thanK method. K;ess thanK method. $n the Kless thanK method we start with the upper limits of the classes and go on adding the fre uencies. 3hen these fre uencies are plotted we get a rising curve. K-ore thanK method. $n the Kmore thanK method we start with the lower limits of the classes and from the fre uencies we subtract the fre uency of each class. 3hen these fre uencies are plotted we get a declining curve.

The following fre uency distribution is converted into a cumulative fre uency distribution first by the Kless thanK method and then by the more than method2 M! 4s #=!%= %=!&= &=!'= N#. #. s"+%&$"s ' * M! 4s '=!(= (=!*= *=!+= N#. #. s"+%&$"s %= #: %

C+2+,!"i'& F &*+&$)3 Dis" i-+"i#$s M! 4s JL&ss "(!$K %= &= '= (= *= += N#. #. s"+%&$"s ' #= %= '= (: *= M! 4s JM# & "(!$ #= %= &= '= (= *= += N#. #. s"+%&$"s *= (* (= '= %= % =

<rom the above distribution one can read at once the number of students who have obtained marks less than a particular value or more than a particular value. Thus there are %= students who have obtained marks less than '= and (= students who have obtained marks more than &=. 1ometimes instead of writing K;ess thanK and K-ore thanK we write Kor lessK and Kor moreK. The implication is different in the two cases, Thus marks less than %= would exclude %= whereas marks %= or lessK would include %=. 1imilarly, marks more than &= would exclude &= whereas marks K&= or moreK would include &=, 7ne has to be very clear about the object in mind before these terms are used. #%=

U"i,i"3 #. O1i'&s <rom the stand!point of graphic presentation, the ogive is especially used for the following purposes2 To determine as well as to portray the number or proportion of cases above or below a given value, To compare two or more fre uency distributions. Aenerally there is less overlapping when comparing several ogives on the same grid than when comparing several simple fre uency curves in this manner. 7gives are also drawn for determining certain values graphically such a.s median, uartiles, deciles, etc, /espite the great significance of ogives, it should be noted tht they are not as simple to interpret as one may feel and hence the reader must be careful while using them.

I,,+s" !"i#$ ; /raw less than and more than ogives from the data given eblow2 (EB %==() P #.i"s /Rs. L!4(s0 #=!%= %=!&= &=!'= '=!(= (=!*= N#. #. C#s. * : #% #: %( P #.i"s /Rs. L!4(s0 *=!+= +=!:= :=!D= D=!#== N#. #. C#s. #* : ( %

S#,+"i#$ ; ;ess than ogive. $n order to draw less than ogive we start with the upper limit of the classes as shown below2 Crofit less than (Rs. ;akhs) ?o. of @os. %= * &= #' '= %* (= '' *= *D += :( := D& D= D: #== #==

O ive by =*ess than= method.

CA@ C@@ #o. of .os. I@ H@ ?@ A@ @ @ A@ ?@ H@ profits ;%s. *akhs< I@ C@@ CA@

#%#

-ore than ogive2 in order to draw more than ogive, we start with the lower limits of the various classes2 Crofit less than (Rs. ;akhs) ?o. of @os. #= #== %= D' &= :* '= +' (= (* *= &# += #( := + D= %

O ive by =(ore than= method.

CA@ C@@ #o. of .os. I@ H@ ?@ A@ @ @ A@ ?@ H@ profits ;%s. *akhs< I@ C@@ CA@

I,,+s" !"i#$ 2 /raw a percentage curve for the following distribution of marks obtained by +== students in an examination. M! 4s =!D #=!#D %=!%D &=!&D '=!'D N#. #. s"+%&$"s D '% *# #'= %(= M! 4s (=!(D *=!*D +=!+D :=!:D N#. #. s"+%&$"s #=% +# %& %

<ind from the graph (i) the marks at the %=th percentile, and (ii) percentile e uivalent to a mark of *(. S#,+"i#$2 6 percentile curve is a cumulative curve drawn on a percentage basis. 9ence for drawings such a curve three steps are re uired2 <ind the cumulative fre uencies of the given data by the 4less than5 method @overt these cumulative fre uencies in percentage of the total Take these percentages on the N!axis and the variable on the R!axis and plot the various points and join them by straight lines. The curve so drawn is known as the percentile curve. M! 4s ,&ss "(!$ D.( #D.( F &*+&$)3 D '% #%% C+2+,!"i'& . &*+&$)3 D (#

P& )&$" #.& +.&

%D.( &D.( 'D.( (D.( *D.( +D.( :D.(

*# #'= %(= #=% +# %& &

##% %(% (=% *=' *+( *D: +==

#*.= &*.= +#.+ :*.& D*.' DD.+ #==.=

Percenti*e curve
Percenta e of students CA@ C@@ I@ H@ ?@ A@ @ D.E AD.E ?D.E (arks HD.E ID.E

$t is clear from the graph that at %=th percentile marks are &#.( and corresponding to *( marks, the percentile is D=. I,,+s" !"i#$ 2 The following table gives the wages of the workers in a certain factory2 D!i,3 5!1&s /Rs.0 %=!%( %(!&= &=!&( &(!'= '=!'( '(!(= (=!(( ((!*= N#. #. 5# 4& s %# %D #D &D '& D' +& *: D!i,3 5!1&s /Rs.0 *=!*( *(!+= +=!+( +(!:= :=!:( :(!D= D=!D( N#. #. 5# 4& s &* '( %+ ': %# #% (

/raw a histogram and a fre uency curve for the data given above. <ind the number of workers whose wages lie between Rs. (+ and Rs. ++. S#,+"i#$ 2 The histogram of the above data is given. $n order to find out the number of workers getting wages between Rs. (+ and Rs. ++ we will draw a less than ogive. /aily wages (Rs.) ?o. of workers /aily wages (Rs.) #%& ?o. of workers

;ess than %( ;ess than &= ;ess than &( ;ess than '= ;ess than '( ;ess than (= ;ess than (( ;ess than *=

%# (= *D #=: #(# %'( &#: &:*

fre7uency curve
#@@ H@@ E@@ ?@@ B@@ A@@ C@@ @ AE BE ?E EE : a es ;%s.<
(I(
No of !or"#rs b#t!##n 55 $ %% & '515(3)5*&1%0

;ess than *( ;ess than += ;ess than +( ;ess than := ;ess than :( ;ess than D= ;ess than D(

'%% '*+ 'D' ('% (*& (+( (:=

&'(

5%

++

HE

#E

IE

Li2i"!"i#$s #. Di!1 !2s !$% G !p(s 6lthough diagrams and graphs are a powerful and effective media for presenting statistical data they are not under all circumstances and for all purposes complete substitute for tabular and other forms of presentation. The well trained specialist in this field is one who recogni)es not only the advantages but also the limitations of these techni ues. 9e knows when to use and when not to use these methods and from its repertoire is able to select the most appropriate form for every purpose. Iulin has beautifully said, ,Araphic statistics has a role to play of its ownJ it is not the servant of numerical statistics, but it cannot pretend, on the other hand, to precede or displace the latter.. The main limitations of diagrams and graphs are2 They can present only approximate values. They can approximately represent only limited amount of information. They are intended most to explain uantitative facts to the general public. <rom the point of view of the statistician, they are not of much help in analy)ing data. They can be easily misinterpreted and, therefore, can be used for grinding one5s axe during advertisement, propaganda and electioneering. 6s such diagrams should never be accepted without a close inspection of the bonafides because things are very often not what they appear to be. The two!dimensional diagrams and the three!dimensional diagrams cannot be accurately appraised visually and, therefore, as far as possible their use should be #%'

avoided.

Pi)"#1 !p(s /ata recorded " presented on the picture is called pictographs I,,+s" !"i#$ ; -arket shares of 9and!set manufacturers (-arch, %==*) +=> ?okia

# ' + U

% ( : =

& * D U

%=> ;A #=> others

1ubdivided Ear diagram2 find the total of the components and draw simple bar diagram for the total. Fach of the bars is divided into the no. of components added together is proportion. $t helps to compare the components of bars at the same #%(

time.

C! "#1 !p(; D&s"i$!"i#$ .# Ai D&))!$ F,i1("s

Fx. 21how the following information by subdivided bar diagram. (1cale # cm to P #= students) @lass 6 E @ / ?o. of students (( (* (' (= #%* 1tudents passed in Fnglish '' '% %+ %&

CA@ C@@ 8 of students I@ H@ ?@ A@ @ 7lass A

+ai,#d

-ass#d

./a,# 0 On 1 a2is 1 /3 & 10%

> .*ass

Ioint bar diagram2 The separate bars are drawn jointly for each component in the joint bar diagram.

Fx. 2 1how the information given below by a joint abr diagram (1cale # cm P #== women) Billage 6 E @ / F
H@@ E@@ #o. of :omen ?@@ B@@ A@@ C@@ @ A > 7 9i**a e / 8

;iterate women ('= ''= ((= &'= '*=

3iterate women .lliterate women

$lliterate women %&= %D= &(= '#= #D=

Cercentage bar diagram 2 <or precise comparison, percentage of each component is calculated and ti is shown by sub!divided bar diagram. 6ll bars are of e ual heights.

Fx. 1how the information given in Fx. ' by a percentage bar diagram (1cale # cm P #=>) #%+

@lass ?o. of students 1tudents passed in Fnglish > students passed in Fnglish
CA@ C@@ 8 of students I@ H@ ?@ A@ @

6 (( '' ''"(( x #==P :=

+ai,#d

E (* '% '%"(* x #==P +(

-ass#d

@ (' %+ %+"(' x #==P (=

/ (= %& %&"(= x #== P '*

./a,# 0 On 1 a2is 1 /3 & 10%

7lass

> .*ass

Cie diagram 2 $t consists of circle. /ifferent classes are represented by different sectors of the circle. The measures of areas of these sectors are proportional to the given data of corresponding class. -easures of arcs of sectors corresponding to each are calculated. 6 circle of convenient radius should be drawn. $n counter clockwise direction angles of measures of arcs corresponding to each class are constructed as shown in figure.

Fx. /raw a pie!diagram representing the information as shown in the table 9ead 6griculture $rrigation Flectricity $ndustry @ommunication 7ther > %( #( #( &= #= ( Total -easure of arc of sector %("#== x &*=PD== #("#== x &*=P('= #("#== x &*=P('= &="#== x &*=P#=:= #="#== x &*=P&*= ("#== x &*=P#:== &*==

+ther, CI 7ommunication, BH Agriculture, D@

.ndustry, C@I

#%:
.rrigation, E?

8lectricity, E?

($$)

Araphical Representation2 9istogram 2 $t is a special type of bar diagram. Rectangales are drawn with base e ual to the class limits and heights proportional to the fre uencies. These rectangles are joined to each other. ($f classes are not continuous, make them so).

Fx. /raw a histogram for the following information (1cale 2 #cm P (>) @lass > #(!%= '% %=!%( &: %(!&= &( &=!&( %* &(!'= #* '=!'( (

?E ?@ BE .*ass &imits B@ AE A@ CE C@ E @ CE:A@ A@:AE AE:B@

1cale 2 7n N axis # cm P (>

B@:BE

BE:?@

?@:?E

Percenta e

M!(! !s(" ! S&!2,&ss Li2i"&% 1ales H $ncome from 7perations

I@@ #@@ H@@ E@@ ?@@ B@@ A@@ C@@ @ /ec.,@B /ec.,@? /ec.,@E B#D.AE HC#.E? #H@.HE

#%D

?et profit
D?.IE

C@@ D@ I@ #@ H@ E@ ?@ B@ A@ C@ @ /ec.,@B /ec.,@? EA.@I EB.I

/ec.,@E

Easic FC1
BE B@ AE A@ CE C@ E @ /ec.,@B /ec.,@? /ec.,@E C#.HE CI.HH BA.D

(?ine months ended)

#&=

CHAPTER 13 REPORT ARITING 13.0 INTRODUCTION

The results of the research 1tudy need to be formally communicated in the form of research reports. The preparation and transmission of a clear, accurate, and convincing report effectively communicates the research findings. This is important because regardless of the uality of the research process and the accuracy and usefulness of the resulting data, the findings will not be utili)ed appropriately by the decision!makers unless they understand the research reports properly. 6 report should give the client an understanding of the data and conclusions. $t should also convince the client that its conclusions are correct so as to obtain the appropriate action from them. Researchers must also make their presentations technically accurate as well as understandable and useful. 13.1 ROLE OR CUALITIES OF THE REPORT

The report serves three chief functionsJ it organi)es the data, establishes the uality of the research and motivates appropriate action. $t organi)es the data, and analy)es the findings in both a logical and lucid manner thus giving them a permanent form. /ue to its systematic style, it serves as an essential reference for future research on related lines. 1ince many executives cannot easily ascertain the uality of a research design, uestionnaire, or experiment, the report is treated as an index of the researcherKs skill and performance. Thus the uality of the report often becomes a significant parameter of the uality of the research itself. The third and the most important role is the determination of the action taken. Ffficient reports lead to appropriate actions or policies. They inspire decision! makers to promptness.

Fach research project is different, thus the process of communicating the results also varies. /epending on the nature of the problem, the information contained therein, and the individual preferences of those who utili)e the report, the findings may be reported in any or all of these forms. 13.2 TYPES OF REPORTS

B!si) &p# " This is composed of working papers and preliminary drafts of the projectKs findings. $t is prepared by the researchers for their own use. The methodology and data contained in these are used by them for reference or to aid other studies. #&#

R&p# " .# p+-,i)!"i#$ These reports are prepared for articles in trade and professional journals, popular maga)ines, bulletins, or monographs. ?ormally, these are relatively condensed reports. The writer has to determine the character and interest of the audience to be reached as well as the publisherKs policies. T&)($i)!, R&p# "s These reports are usually intended for scientific or technically trained persons. 1pecific descriptions of the entire procedures employed, including introduction of the problem and the hypotheses researched, details that lead to conclusions, complicated technical appendixes on the methodology and complete bibliographies are provided in the report. R&p# "s .# &<&)+"i'&s These reports contain major conclusions and recommendations of the research project and are intended for decision!makers. Mnlike technical reports, voluminous details and methodological information donKt receive much focus here (they are generally put in an appendix, so that they can be referred to if the need arises). This is so because not only are managers extremely busy, they also lack the skill and interest to understand the technical und logical aspects and terminology of research. 6s findings can be presented orally or in writing, the term LreportL refers to either form of presentation. @lients generally appreciate an oral presentation from the researchers in addition to the more essential written reports, as it provides an opportunity to clarify their doubts. 9ence the ability to communicate well in both speech and writing is essential for the researchersK success. 13.3 PRINCIPLES OF REPORT ARITING OR STEPS IN REPORT ARITING

The fundamental form of communicating research findings is through written reports. The entire information that has been gathered during the research process now has to be passed on to the decision!makers so that they can take appropriate actions. Fach report has to be tailored depending on the needs and profile of the users. $f the report writer knows the tastes and thought modes of the client managers, then the presentation is likely to be more effective. 7therwise, failure to understand the nature and capacity of these individuals, their interest, or lack thereof, in the subject area, the circumstances under which they will read and evaluate the report and the uses they will make of the report may negate the very purpose of the report writing. Fxecutives have their own preferences. <or instance, some might prefer a short report, only on results. 7thers may demand the research methodology and results together in the report. 1ome may want considerable information on the research methods used in the study. -any #&%

executives emphasi)e brevity while others demand complete discussion. Thus the audience determines the type of report. 1o researchers have to be aware of such wide diversity of interests and preferences and accordingly choose to present their study report. Aiven below are some specific pointers as to how written reports have to be made to suit executive re uirements. F#)+s #$ "(& O-8&)"i'& #. "(& S"+%3 $n general, the research is designed such that the information gathered facilitates a decision to be taken regarding some issue. 9ence the report should be built around this decision and on the research findings that are relevant to it. Us& #. $#$ "&)($i)!, ,!$1+!1& The report should be aimed at the experience level of the reader. 1ome researchers are under the false impression that they can convince management of their expertise and thoroughness in research by giving technical details. Technical details might be of use to some readers, but not to all. 1o it is better to replace technical terms with descriptive explanations. M!4& i" &!s3 "# .#,,#5 The material should be structured logically, the body of the report should be self!evident and the topics easy to find. <or instance, explicit headings indicating every different subject and subheadings for subtopics make it easy for the busy managers to skim through the report and get the gist of it. 6lso certain subtle aspects such as short and focused paragraphs make the report easy to follow. M!4& i" ),&! @larity in writing is an essential uality because vagueness can produce wrong decisions and substantial losses. ?ot speed but clarity should be the aim of report writers. 1imple and familiar words should be used rather than difficult and complex words. 1ince slang and clichgs can be misinterpreted they are best avoided. The report should have a uniform style and format, thus making it easy to read. Us& 1##% s&$"&$)& s" +)"+ & -ost skilled writers use short sentences as long, involved sentences are difficult to read. Coorly constructed sentences lead to confusion, whereas well!constructed sentences help the reader think clearly. 6voiding errors in grammar and spelling is extremely important. This is because one incorrect sentence or misspelled word can undermine the credibility of the entire research project. M!4& i" i$"& &s"i$1 "# &!% #&&

Research reports should be interesting to read. They should not be dull, tedious, or boring. -ere recording of facts without purpose or organi)ation inevitably results in confusion that leads to loss of interest. Fach paragraph should be written with an awareness of its position in the entire report. M!4& i" - i&. !$% -& s&,&)"i'& The report should be just long enough to cover the objectives of the study. $t should highlight major points by stressing the big issues and taking them up first. @omprehending significant points and writing concisely and to the point brings in clarity and impresses the reader. $t is important to exclude anything unnecessary. Eut this is entirely dependent on tile judgement of the research writer as it is difficult to decide to what extent the explanatory material has to be used. Fxplanation of technicalities and the logic underlying the conclusions can either confuse or clarify the reader depending upon how skillfully they are presented. 7ne solution is to mention that certain details are omitted but are available on re uest. S" &ss p !)"i)!, !)"i#$ There should not be an impression that the findings of the research are true only theoretically under ideali)ed conditions but not in reality. Msing analogies, scientific examples, or comparisons drawn from experienceJ familiar to the reader increases the validity of the report. V! 3 "3p#1 !p(3 Bariations in type si)es and skillful use of white space will attract attention and facilitate reading. Mse of capitals to emphasi)e central points, use of uotations, italics, or underlining, dots, exclamation marks, and lead lines direct the attention to significant parts of a page. Us& #. 'is+!, %&'i)&s i$ "(& &p# " Msage of graphs, maps, pictures gives reports a dynamic uality and emphasis. They should be used to supplement and not replace the written text. 6s a general rule, a sentence in the text of a report should not contain more than two or three numerical values as they make sentence difficult to read and understand.

13.4

THE REPORT FORMAT OR LAYOUT

6 report must use the format that best fits the needs and wants of its readers. Though there is no one best format for all reports, a basic outline is provided below that gives a picture of the structure of the report. The following items represent the conventional and logical arrangement of the steps involved in report preparation. #&'

#. %. &. '. (. *. +. :. D. #=. ##. #%.

;etter of transmittal accompanying the report Title page Table of contents Fxecutive summary $ntroduction 1tatement of objectives -ethodology ;imitations <indings @onclusions and recommendations 6ppendix Eibliography

#&(

CHAPTER 14 BIVARIATE MEASURES OF ASSOCIATION 14.0 INTRODUCTION

-arketing -anagers are very often re uired to find the degree of association between variables. <or example, they may need to know the association between the sales and price of a product or the effect of an advertisement on sales. 1uch association can be measured by the use mathematical methods. This chapter deals with only those methods that measure the association between two variables. $n the next chapter we shall discuss techni ues that can be used on many variables. Eivariate methods that use ratio ad interval data will be discussed in detail. 3e will also examine in brief bivariate methods involving nominal and ordinal data. <or each of these methods we will measure two components 2 one, the nature of association, and two, the strength of the association. The nature of the association will allow us to predict the dependent variable at any given level of independent variable. The strength of association measures the degree of association i.e. the variation that can occur in actual value from the predicted values. Barious assumptions are made for each techni ue and these are mentioned in the appropriate sections. 7ne should understand these assumptions thoroughly to correctly interpret the results of the analysis. 7therwise one would make erroneous interpretations of the analysis. Two common errors made in the use of methods of association are 2 (#) association represents causation and (%) association of unrelated variables. $f two variables show association, it only means that they occur together, it does not necessarily mean that one causes the other. ;et us assume that an association between sales of pens and sales of student notebooks was established. This does not mean that the sales of notebooks affect the sale of pens. $nstead, these associations occur due to seasonal variations in the demand for student stationery. 1imilarly, one should check if the relation is of practical value. <or example, though there may be a statistically significant correlation between internal employee turnover and sales figures, the association is not really a practical one. To take a more exotic example, one may find a correlation between rainfall in @herapunji and the sale of shoes in 1yria. 1uch correlation is possible due to the nature of statistics.

P &'i#+s P+ )(!s&

I$"&$"i#$ "# -+3 i$ .+"+ & Y&s N# #&*

T#"!,s

Nes

'= #=

?o Totals

%= (**.+>) @ &= (%(>) (+(.=>) (= (=

E *= (&&.&>) @ '=

(#==>) (#==>)

@alculation of Chi (h) @oefficient 14.1 CROSS TABULATIONS

The simplest way to look for associations is to cross tabulate the data. This method is also referred to as gross break. $t is usually used for nominal data though it is also applicable to interval or ratio data. This method is the lest complicated and the most useful with nominal data. Bariable N Eo Ee Bariable R /o /e @o @e Total Eoe/o 6oe@o 7 subscript means observed fre uency and F subscript means expected fre uency. Fxpected fre uencies are calculated by the formulae E@ Q 6/ P !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! (6e@) (Ee/) (Ee6) (/e@) ;et us look at the following example. The table depicts two variables 2 previous purchase experience and intent to purchase. Eoth are nominal variables. The sample si)e is #==. The table represents the values both as absolutes and as percentages. The percentages are calculated along rows. Msually, the independent variable is positioned in the left margin (the rows) and the dependent variable is positioned in the top margin (the columns). The extreme right column gives the number of people (in this case *= out of #==), who have previous purchase experience. 7f these, '= people (**.*+>) intend to purchase the product again while %= people (&&.&&>) #&+ ? 6o 6e /oe@o Eoe6o

have no intention of buying the product in the future. $f there are relatively high percentages on any diagonal it indicates strong relation. $n the example we find that more people with previous purchase experience wish to repurchase, and those who did not purchase before do not wish to do so.. This indicates that repurchase is strongly and directly dependent on previous purchase. $f the other diagonal had high fre uencies it would have indicated an inverse relationship. Bariable N E @ 6e@

Bariable R Total

E / Ee/

Ee6 /e@ ?

Cast Curchase Fxpenditure C#$"i$1&$)3 !$% P(i C# &,!"i#$

%= &= (=

$ntention to buy in future '= #= (=

*= '= #==

3e often need to establish the strength of a correlation in more uantitative terms than those used above. The two coefficients h(Chi) and @ (contingency) can indicate the strength of a relation. The calculation of these coefficients is given below. The range of values the variable can take is Q # to #. 6s it approaches #, it shows a stronger relationship. h (Chi) P !=.'# @alculation of @ (contingency) @oefficient (Eo Q 6o) (6o e @o) 6e P ? (Eo e 6o) (Eo e /o) Ee P ? (/o e @o) (6o e @o) @e P ? (/o e @o) (Eo e /o) /e P ?

Cast Curchase

%=

$ntention to buy in future '= #&:

*=

Fxpenditure &= Total ?ow (o!e)% _% P ` !!!!!!!!!!!!! & @P _% !!!!!!!!!!!!!! _% e ? (=

&= #= %= (=

&= '= %= ?

Msing the above formulae we get @ P =.&: @ can take values from = to #, but will never approach #. The closer the value to #, stronger the relation. S)!""& %i!1 !2s $nterval data can be correlated in many ways. <or better understanding, data should first be plotted on a graph. $n <ig.#'.#, three graphs are shown. @ase 6 shows a graph of highly correlated data, @ase E displays highly 3here dispersed data, and @ase @ shows a non!linear relationship.

&( &= %( %= #( #= ( = = ( #= #( %=

&( &= %( %= #( #= ( = =

( #= #( %=

&( &= %( %= #( #= ( =

#&D


= ( #= #( %=

Fi1. 14.1; S)!""& Di!1 !2 1catter diagrams serve the following purposes. They 14.2 visually give an idea of the degree of correlation (for instance, the Fffect of advertisement on sales). distinguish a non!linear correlation from a linear one (for instance, ;earning curve) $dentify some extreme data points which may interfere in the analysis. REGRESSION ANALYSIS

Regression analysis is based on the relationship between two or more variables. Through regression analysis, the researcher attempts to predict the value of an unknown variable based on past observation of that variable and others. The variables involved are both independent (known) and dependent (unknown). $n regression analysis, there will be only one dependent variable, whereas the number of independent variables can be more than one. 6s the number of independent variables increases, the accuracy of the predictions will also increase. 6 causal relationship between the variables (independent and dependent) can often be found because changes in independent variables may cause a change in the dependent variable. 9owever, it will not be true to always assume that the independent variable has caused the dependent variable to changeJ the change may be due to another variable that we have not considered. 9ence, it becomes important for us to consider the relationships found by regression as relationships of association. $t need not be a cause and effect relationship. The regression line is a line that best describes the nature of the data presented. $t can be linear or non!linear depending on the relationship between the variables. 3hen the relationship between two variables is linear, the resulting line is linear. To estimate the linear regression line between two variables, researchers need to use the following formulae. The e uation of any straight line is given by the e uation N P a e bR #'=

3here R and N are variables and a is the N intercept of the line and b is the slope. The lines are different from each other only due to the different values of a and b. 9ere N is the dependent variable and R is independent variable a and b are calculated as follows 2 a P N Q bR bP `RN ! n`RN `R% Q ?R%

Msing this regression line a researcher can predict the values of the dependent variable for any given value of independent variable. To find the validity of the estimated line we can calculate a standard error from the following e uation. 1P ` (N!N%) !!!!!!!!!!!!! n!%

N P Balues of dependent variable N P Fstimated value from estimating e uation that correspond to each N value ? P number of data elements used in fitting the regression line. There is a shortcut formula to calculate the standard error2 N P Balues of dependent variable R P Balues of independent variable 1e P ` N% Q a ` N Q b `RN !!!!!!!!!!!!!!!!!!!!!!!!!!!!!! n!%

6 P N intercept calculated using the e uation given above. b P slope of the e uation as given in the above e uation. The larger the standard error, the greater is the scattering or dispersion of the data. Msing this error, we can calculate the interval estimate at a given confidence level for a N value calculated for a given x. The method for finding a confidence interval is similar to the method for predicting a regular confidence interval. 14.3 CORRELATION ANALYSIS

3e may often need to know the degree of linear relation between two variables. <or instance, when we have two variables, advertising expenditure and sales revenue, we may want to know to what extent there exists a correlation between the two. This can be found by using #'#

the correlation coefficient. 3e have two measures for describing the relation Q coefficient of determination and co!efficient of correlation. T(& C#&..i)i&$" #. D&"& 2i$!"i#$ This is the primary way to measure the extent or strength of the association that exists between two variables. $t is also referred to as sample coefficient of determination if it is used on samples, ad it is symboli)ed by r%. $t is given by the following formula2 r% P # P `(N Q N%) `(N Q N)%

The value of r% can vary between = and #. $f the value of r % is e ual to #, it means that the regression line is a perfect estimate of the data. $f the value of r % is e ual to =, it means the regression line is a totally inaccurate estimate of the data. 1tatisticians interpret the coefficient of determination as the amount of variation in N explained by the regression line. $f the value is e ual to #, the variation of N is fully explained by the regression line. $f the value is =, the regression line is totally unable to explain the variation in N. 3e have a shortcut formula for finding this coefficient. R% P 3here r% P 1ample coefficient of determination a P N intercept b P slope of best!fitting estimating line n P number of data points R Pvalues of the independent variable N P values of the dependent variable N P mean of the observed values of the dependent variable. C#&..i)i&$" #. C# &,!"i#$ This is the second measure for finding the strength of the correlation between two variables. $t is actually the s uare root of the coefficient of determination and is represented by r. The values of the coefficient fall in the range of !# to #. 9ere again )ero represents poorly explained correlation. # represents a strong inverse relation between the variables. 14.4 ERRORS POSSIBLE IN THE USE OF REGRESSION AND CORRELATION ANALYSIS a`N e b`RN e nN% `N% Q nN%

The following avoidable errors are commonly made in using these techni ues 2 #'%

 14.6

Researchers tend to extrapolate the data beyond the range of observed data They tend to associate a cause and effect relationship between the Bariables as explained before. 9istorical data is often used to predict future trends. 9owever, in such 1ituations, the data should be current. The conditions during the prediction Ceriod should not be different from the conditions during data collection. Researchers tend to read the values of r and r% as percentages of correctness of the estimation. 9owever, this is not a correct reading. Researchers also find relations when they do not exist, as already explained. LINEAR DISCRIMINANT ANALYSIS

Regression techni ues are useful in analy)ing intervally!scaled data. $t is often necessary in research to establish relationship between nominal variables, i.e. will buy Q will not buy, heavy user Q light user, credit purchase Q cash purchase, etc. $n such situations, /iscriminant 6nalysis is useful. /iscriminant analysis classifies people or objects into two or more categories using intervally 1caled predictor (independent) variables. $t can also be used to generate perceptual maps. The mathematical logic is similar to regression analysis. $t is too complicated to be computed manually and is computed by computer. The programs produce n!# formulae for a problem with n categories. <or example, if we have six categories like very high user, high user, medium user, low user, very low user and on!user, then for these six categories, we obtain five formulae. 3e use these formulae to classify the objects or persons. 1ome computer programs produce two formulae, while some others produce one formula for two!category analysis. ;et us take a look at an example in which one formula is produced. Eefore proceeding with the analysis, we need to check whether the programs are able to give the desired level of significance. 3e should proceed only if this condition is met. 6ssume that we obtained the following formula for a problem. N P !=.#R# e =.(R% Q =.*(R& 3here N P Msage of @able TB R# P The si)e of the family R% P Cresence of college going children R& P Cresence of retired people in the family 3e have to follow the steps given below to find a critical value that is useful in classifying the consumers in future. #'&

#. %.

The means of each of the independent variables are calculated separately for user and non!user. The means of user groups are used to produce a value for N users while the means of nonuser groups are used to produce a N value for nonusers.

The critical value is the midpoint between these two N values. 3hen there is new data, one substitutes the values of dependent variables in the e uation to obtain the N value. $f the N value is higher than the critical value, then the family is a user, else a nonuser. The accuracy of the discriminant analysis is tested by the use of the classification matrix, also known as the confusion matrix. 9ere a sample of known users and nonusers is classified by using the above critical value. The sample can be nothing but the original sample. Eut using the same data to generate a value and then using the same sample to test it may bias its accuracy. 1o, many a researcher holds back a part of the original data from the discriminant analysis and then uses it for testing. 6 confusion matrix for a purchase intent problem is shown in Table #'.#. The number of correct and incorrect classifications is indicated by the percentages in brackets. Crevious Curchase Fxperience P#si"i'& $ntend to buy in future /o not intend to buy Total '(= (+(>) #(= (%(>) *== N&1!"i'& %== (%(>) *== (+(>) :== T#"!, *(= +(= #'==

14.@

AUTOMATIC INTERACTION DETECTOR

The best way of staying ahead of the competition is by discovering information unknown to the competitor. Today5s companies have huge databases that contain a large data. 6 huge amount of data from past research is also available. 9owever, this data is not checked for all possible associations. This is so because it does not make sense to make some relations. Eut some of such irrational relationships have worked out all over the world. <or example, a super market chain found a relationship between sales of beer and diapers. Ey using this knowledge they could sell more by placing these goods together on the shelves. 1uch unknown relationships can exist between any dependent variable and some independent #''

variables. This can be detected by the use of a techni ue called 6utomatic $nteraction /etector 6M/), $t us possible to even find a relation between a set of dependent variables and a set of independent variables. This method is made more efficient by the use of computer. Easically, the 6$/ techni ue is to systematically breaks down a large group of sample units into smaller subgroups that share similarities in terms of dependent and independent variables. The measures of associations in this chapter are useful for finding relationships between various variables. These relationships are useful for predicting the values in future. Eut these techni ues are useful only if one knows their limitations and uses them appropriately. Mnfortunately, these techni ues can be manipulated or misapplied. $f a researcher or a manager wants support for his decisions, it is possible to achieve this by appropriate manipulation of the data. Eut this does not mean that one should condemn the techni ues. $nstead it is the attitude of the people who use it that is at fault. 7ne can even get wrong results by using these methods inappropriately. APPENDIB ; LIST OF STATISTICAL PACLAGES <ree 1oftware #. %. &. '. (. *. +. :. D. #=. ##. #%. #&. #'. #(. #*. #+. #:. #D. %=. %#. %%. %&. %'. %(. 6/F!' 6nderson 1tatistical 6rchives 6R$-6 E"/ Eoomer /F 9istograms Fasy-6 Fasy1tat Fmpty @orners Fpi$nfo <irst Eayes AUCower ArafCrog $nsight $nstat I/E ;7@<$T -dac6nova -6?FT -eta!analysis (.& '. 6nalyse!it for -icrosoft Fxcel (. 6nswer Tree %.= *. 6pplied 6nalytic 1ystems +. 6ssi1tat :. 6MT7E7R D. E-/C #=. Eilog & ##. @hemetr(ica #%. @linstat #&. @o9ort #'. @rossgraphs #(. @rystal Eall #*. /ata /esk #+. /E-1"@opy #:. #D. %=. %#. %%. %&. %'. %(. %*. %+. %:. #'( Fa 1t Fasy1tat Fc1tatic Fgret Fpi $nfo FC$-FT6 F01 F uivtest Fviews Fxecu1tat <orecast Cro

%*. %+. %:. %D. &=. &#. &%. &&. &'. &(. &*. &+. &:. &D. '=. '#. '%. '&. ''.

-eta!analysis calculator for 2 %D. -urdoch, /uncan. &=. C6-@7-C Regress Rweb 16ERF 1cilab 1erpik Araph 1$16 1tatistical 1oftware by Caul 3. -ielke Ir. 1TC;6? Bi1ta 3eb1tat 3eighted ;east 1 uares ;inear <its 3in166-

@ommercially 6vailable 1oftware #. 6ctivstats %.6mos '.= &.6naly @orp

<orecast R A6M11 &#. AE!1T6T &%. Aen1tat &&. A;$&'. $nsight &(. $nstat &*. $nterleap &+. I-C &:. ;imdep &D. ;ispstat '=. ;$1RF; '#. ;ogRact '%. -ac6nova '&. -aestro ''. -aple '(. -atrex '*. -esa '+. -initab, $nc. ':. -;6E 'D. -7/1T6T (=. -ultivariance +.&% (#. ?@11 (%. ?omparametic bootstrap regression (&. ?C1T6T"?C<6@T ('.n0wery 6dvisor ((. 7riana (*. C!1tat (+. Cermustat

(:. Crism (D. Croceed *=. Crostat *#. C1$!Clot *%. 0ualitek!% *&. R6T1 *'. Resampling 1tats. *(. %Risk (1preadsheet 6dd!in) **. 1!Clus *+. 161 *:. 1ha)am *D. 1cientific 1oftware

:%. 1tatistix :&. 1tatit :'. 1tatlab :(. 1tatmost :*. 1tat Biew :+. 1tatRact ::. 1T6TBFR :D. 1M/66? D=. 1N1T6T D#. Testimate D%. TC; Tables D&. T1C #'*

+=. 1igma 1tat +#. 1$-1T6T for 3indows +%. 1olas +&. 1oritec for 3indows +'. 1C11 +(. 10@pack +*. 1T6T6 @orp. ++. 1ta Table +:. 1tat Fase $nc. +D. 1tatAraphics :=. 1tatistica :%. 1tatistical @alculator.

D'. Mni1tat D(. Bitalnet D*. Bi)ion D+. 3esBarC@ D: 3$?G1 (3indows G3$1T6T) DD. 3in1T6T #==. xlstat #=#. RCro #=%. 1C11

#'+

CHAPTER 16 MULTIVARIATE MEASURES OF ASSOCIATION 16.0 INTRODUCTION

3e learned about measures of association between two variables in the previous chapter. $n this chapter we shall discuss measures of association between more than two variables. $t should be noted that the same precautions one takes with bivariate methods are necessary here also. To assist recall, let us repeat the main points. 6ssumptions underlying each of the methods should be fully understood and utili)ed to obtain accurate results. 6ssociation between two variables does not imply a causal relationship. 6 mathematical relationship may occur due to the very nature of randomness in sampling, but experience and judgement should be used to accept or reject a relation. 16.1 MULTIPLE REGRESSION

$n the previous chapter we saw how Regression 6nalysis is used in marketing research. 3e also mentioned that more than one independent variable can be used to predict the dependent variable. This process of using more than one independent variable to predict the defined variable is called multiple regression. ;et us take a case with three variables. 1uppose we want to correlate the sales data with the number of advertisements and the rise in per capita savings. The dependent variable, sales, can be represented by N, and the independent variables, per capita savings and number of advertisements, can be represented by R# and R%. Table #(.# gives the data collected. 6s we have seen, there is a general e uation for a line. 6 line with n independent variables has the following general e uation. N P a e b#R# e b%R% e b&R& e ... e bnRn 3here N P dependent variable R#, R%, ..., Rn P independent variables a P N intercept b#, b% ..., bn P slopes associated with R#, R%, ...,Rn respectively <or a two independent variables problem, as in the above case, the e uation reduces to P ae b#R# e b%R% Balues of a, b# and b% are to be determined. ?ow we obtain e uations known as normal e uations, the values of which are readily available. <or three `N P na e b#`R e b%`R% variables the e uations are solving these e uations we find the values of a, b# and b% #':

`R#N P a`R# e b#`R#% eb%`R#R% `R%N P a`R% e b#`R# R% eb%` R#% a b# b% !#%.&:(D%#++ #.#=.++=DDDD !=.=%'(::&%D

7nce we find these values, we can forecast any value of dependent variable for given values of independent variables. To find the accuracy of this estimate we need to find the standard error of the estimate. This 1e P `(N! i)% !!!!!!!!!!!!! nQkQ#

can be calculated by the following formula 3here N P 1ample values of the dependent variable P corresponding estimated values from the regression e uation n P number of data points in the sample k P number of independent variables Regression values for the problem at D= percent confidence -ultiple R R 1 uare 6djusted R 1 uare 1tandard Frror Y&! #DD' #DD( #DD* #DD+ #DD: #DDD %=== =.D*&%#D#D =.D%++D#%# =.:D#*:*:#( &.%&:#+='(%

S!,&s i$ N+2-& #. P& )!pi"! &! $i$1s "(#+s!$%s !%'& "is&2&$"s "(#+s!$% +p&&s '( #%= &+ &(.( ##% %& (* #'& '( *' #:= (* (' #(* &D ': #': '= (% #(+ '# T!-,& 16.1; D!"! .# M+,"ip,& R&1 &ssi#$ P #-,&2

Us& #. M+,"ip,& R&1 &ssi#$ i$ M! 4&"i$1 R&s&! )(

#'D

-ultiple Regression can be used for solving a variety of marketing problems. 1ome of them are listed below2 #. Forecasting where either company variables (relative price, relative advertising etc) or external variables (population growth, disposable income etc) or both are used to forecast sales, demand, or market share. Outlet location decisions where the traffic, the location of competitors, the s uare footage available, and so on, are used to analy)e the attractiveness of outlet locations for chain stores. Quota determination involves using territory si)e, last period sales, competitor strength and related variables to determine sales uotas or objectives. Marketing mix analysis by analy)ing the relationship between elements in the marketing mix and market share or sales. Determining the relationship between the criterion variable and one predictor variable while the effects of other predictor variables are held constant. 6n example is the estimate of the reliance on price as an indicator of the uality of furniture, while other factors, such as the brand of the product and the stores in which it is available are held @onstant. Fstimating values for missing data (item non! response) in surveys. <or example, one can estimate income from occupation, age, and education data.
<ormal Fducation 7n!the!Iob Training

%.

&. '. (.

*.

Trainin g

Cerformance

-anager5s 7bservation 3ork 7utput -easurement

-otivation Test

Eehaviour 7bservation Lis &,

-otivati on

;$1RF; is an acronym for ;inear 1tructural Relationships, and it was first introduced by Ioreskog in #D+&. $t is a highly complicated techni ue, but as it is extremely useful in explaining causality among constructs that cannot be directly measured, we shall discuss it here. #(=

;$1RF; analy)es covariance techni ues through two models. The first one is a measurement model and is used to relate the observed, recorded or measured variables to the latent variables (constructs). <or example, when we are measuring an employeeKs perfom#a/ce, a construct, we relate it with variables like ! (#) managerial observations (%) recorded measures of output (&) performance appraisal by peers and (') performance appraisal by subordinates. These variables together give the researcher an understanding =# the employeeKs performance. This type of measurement model for useful in modeling constructs like feelings, performance, behavior and attitudes. The second model is a structural e uation model. This model shows the causal relationships among the latent variables and also describes unexplained causal effects and variance. $t consists of a set of linear structural e uations. <or better understanding, it is often diagrammed by the use of path analysis and the resulting diagram is called a path diagram. <igure #(.# depicts one such path diagram for employee performance. The circles represent latent variables, the rectangles the measurable variables while the arrows represent the causal relationships. The latent variables give rise to the measurable relationships as indicated. The arrow between KmotivationK and KtrainingK indicates the interrelationship between these variables. A"" i-+"& @olour Crice Fxternal style and /esign D&)isi#$ ,&'&,s 1ilver 1ky Elue -aroon Rs.(,==,=== Rs.',(=,=== Rs.(,(=,=== 6lpha Eeta Aamma /elta 9igh -edium ;ow : G-";itre #= G-";itre #% G-";itre #' G-";itre

Cerformance levels -ileage

T!-,& ; 16.2 ; A T!-,& #. A"" i-+"&s .# C#$8#i$" A$!,3sis

16.2

CONJOINT ANALYSIS

3henever there are some attributes that the consumer is ready to trade off or compromise on, researchers use conjoint analysis to identify the products that meet their needs best. @onjoint #(#

analysis is used with non! metric (non! uantifiable) independent variables. 1uch analysis is often used in new product development and for positioning. Msually, cross tabulation is used when the variables are two or three in number. Eut as the number of attributes grows, the problem becomes too complex to be handled by cross tabulation. <or instance, if we are testing for an ideal new automobile, we can have three colors, three prices, four external designs, three performance levels and four mileage factors. The total decision levels in the model are (&R&R'R&R') '&%. $t is not possible to make so many cross table comparisons. @onjoint analysis is advisable under such circumstances. M&"(#% The attributes most pertinent to the study are first selected by using the experience of the researcher or an exploratory study or a combination of both. The above automotive example is represented in detail in Table #(.%. The researchers prepare cards with combinations of various attributes. These combinations are later given to the consumers for ranking. <rom these railings one can infer the relative importance of each attribute. This is also known as part!worth or utility scores. B#< 16.1 C!p"+ i$1 C+"#..s $t is not possible to find the customer response to each of the '&% combinations in this 3e can Researchers elicit cutoffstherefore with simple uestions like these2 problem. use a minimum set of configurations from which the response the most you consider paying forThis a car8 VSSSSSSSS 8 usually generated by a to3hat each is attribute can bewould independently assessed. minimum set is L3hich of the following B@R brands would you never consider buying8 (check all that computer package. The cards can be as given in Table #(.& apply) jk-agnavox 7nce the rankings are obtained, the results are again fed into a computer software and the j k 1ony results are analy)ed. j k 9itachi j k 1arnsung CAR A /C! % 10 CAR B /C! %20 1ilver 1ky blue ThatKs all there is to it. 0uantitative attributes (like memory, hard drive si)e, @/!R7Rs.(,==,=== Rs.',(=,=== speed, or processor speed for C@s) can be asked as in the first uestion. 0ualitative 6lpha Aamma attributes (for C@s, type of storage drive, C@ brand, processor brand, or menu of pre! 9igh -edium loaded software) can be asked as in the second uestion. #= G-";itre : G-";itre I$)# p# !"i$1 C+"#..s T!-,& 16.3; S!2p,& C! %s .# C#$8#i$" A$!,3sis i$ A+"#2#-i,&s 3e can incorporate these cutoffs into our conjoint models as penalties. 1ay a car is V&=,=== and Iones has a cutoff of V%(,===, but 1mith does not. The model produces a coefficient, or utility, for the price of V%(,===. $t also identifies a penalty for the fact that 16.3 FACTOR ANALYSIS the price is V(,=== over IonesK cutoff. The penalty impacts Iones, but not 1mith. The inclusion of this information about individualKs idiosyncratic tastes should improve the The term factor analysis was first introduced by thurstone in #D&#. The main applications of predictive ability of the model, and to the extent it does, it is beneficial to include. this techni ue are2 (#) to reduce the number of variables to a manageable number and (%) to detect the structure in the relationships between variables (that is, to distinguish variables T(& B#""# ! Li$& from variables that belong together and have overlapping measurement characteristics). $n recent studies we tested the value of including cutoffs in conjoint analysis. Eoth studies included LholdoutL choice uestions whose #(% purpose was to test the accuracy of alternative predictive models. -ean absolute deviation (-6/) is a measure of how far off, on average, are a modelKs predictions of holdout choice shares. 1mall -6/ is good and big -6/ is bad.

$n addition to the more accurate simulators we can construct by including the cutoffs, we can also report the incidence and magnitude of these cutoffs. This extra information adds a nice table to our reports. 9ereKs an example of part of the cutoff information from the panel study of C@ attributes2 A"" i-+"& ? L&'&, %&& -h) 6-/ processor %** -h) Centium $$ &== -h) Centium $$ @ost 2 V #%== @ost V#'== @ost V#*== U"i,i"3 !.*# !=.% .*& .&& .=+ !'= M 5i"( )+"#..s %# #% #= ( #% %# C+"#.. P&$!,"3 !*.( !.'D ! ! !.'D !.D+

#(&

This table shows that the utility of a %** -h) Centium $$ is .=% for most consumers, but

$t is often possible that the data collected in a study may be reducible to a lesser number of variables. 1uppose we conduct research in which one uestion measures the current satisfaction of the consumer using a product, while another uestion measure the repurchase fre uency. $t is often possible that both these variable are highly correlated. This means that one of the variables is redundant and so can be deleted. $nstead, a regression line between these two is calculated and is used in place of thee two variables. This method can be used only if the correlation is highJ it can also be extended to multiple variables. M&"(#% <actor 6nalysis can be done in many ways. 3e shall take a look at the most fre uency used approach known as principal component analysis. $n this method, the given set of variables is reduced to another set of composite (combination of one or two) variables are not correlated to one another. These new variables, known as factors, account for the variance of the entire data. E<" !)"&% F!)"# s $ $$ $$$ $B V! i!-,& 6 E @ / F < Figenvalue Cercent of variance P& )&$"!1& #. "(& &2!i$i$1 '! i!$)& "(& .!)"# !))#+$"s .# (*> %&> #*> '> U$ #"!"&% F!)"# s .+= !.'= .*= !(.= .*= !&.( .(= .(= .*= .(= .*= .*= %.#: #.(D &*.&= %&.%= #(' C+2+,!"i'& . &*+&$)3 (*> +D> D(> DD> R#"!"&% F!)"# s =.+D =.#( =.+( =.=& =.*: =.# =.=* =.+ =.#& =.++ =.=+ =.:( #.*D #.D' =.%: =.&%

@umulative variance

&*.&=

(D.(=

=.%:

=.*=

T!-,& 16.6 ; F!)"# 2!" i)&s The best combination of variables forms the first factor and accounts for most of the variance. The next factor is chosen such that it is the best combination that accounts for variance unexplained by the first factor. This process is repeated until all the variance is accounted for. Eut practically, a researcher stops after most of the variance is accounted for. Table #(.' represents the probable data obtained while performing a principal component analysis. =.: =.* =.' =.%
Rotated <actor $$

!=.% !=.'

=.%

=.'

=.*

=.:

!=.* !=.:

Rotated <actor $

$ncome

@ #((

6ge Results from a factor study are given in Table (adapted from @ooper and 1chindler). The values in this table are correlation coefficients between corresponding factors and variables. <or instance, =.+= is the correlation coefficient r between variable 6 and factor $. These coefficients are called loadings because they indicate the amount of variance of the variable each factor is loaded with Figenvalues are the sum of s uares of variances of factor values. <or instance, the Figenvalue of factor $ is (=.+=) %e(=.*=)%e(=.*=)%e(=.(=)%e(=.*=)%e(=.*=)%. 3hen the Figenvalues are divided by the number of variables (here *), we get the total variance explained by the factor. $n this table we note that the loadings between factor and % are not mutually exclusive. 3e want the two factors to have as little correlation between them as possible. This means that the loadings should not overlap. $deally we should get loadings such that each factor is loaded only in those variables in which the other variable loadings are )ero. Eut we see that the unrotated factors have high loadings on /,F and < in both the factors, indicating a correlation between the two factors. To correct this situation there is a method known as Rotation of 6xes. This can be orthogonal or obli ue. To understand orthogonal rotation, assume that we are dealing with a two dimensional space rather than a multi dimensional one. The values obtained as produced in Table are shown in. Two axes are used here. Eut the location of these axes is only arbitrary and one can rotate the axes without changing the real loadings. 7nly the relative values of the loadings change. ;et us rotate to obtain better values for the factors. The rotation resulted in the factors shown in Table #(.( in the rotated factors section. ?ow we see that the loadings are less correlated with each other. <actor $ is highly loaded in variables 6, Eand @ while <actor $$ is highly loaded in Bariables /,F and <. Thus one can obtain the best factors by using this method.

16.4

CLUSTER ANALYSIS

6 general uestion facing researchers in many areas of in uiry is how to organi)e observed data into meaningful structures, i.e. taxonomies. @luster 6nalysis (first used by Tryon, #D&D) was developed to cater to such a need. 3henever one needs to classify a LmountainL of information into manageable meaningful piles, cluster analysis is used. Mnlike discriminant analysis, it does not have an a priori hypothesis. /iscriminant %nalysis begins with a well! defined group composed of well!defined groups with two or more distinct characteristics, and tile analysis provides a set of variables to separate them. Eut cluster analysis starts with an undifferentiated group of people, events and objects and reorgani)es them into homogenous groups. #(*

M&"(#% 7nce the sample is selected the following steps are gone through to arrive at clusters Bariables on which objects, events, or people are to be measured should be defined. The variables can be financial status, si)e of the family, education, etc. The data for these variables should be collected. 1imilarities among the elements, persons, events or objects should be calculated by the use of correlation, Fuclidean distances, and other techni ues. -utually exclusive groups, called clusters, should be made. To achieve this, within!cluster similarity and between!cluster differences should be mathematically maximi)ed. 6lternatively one can arrange the clusters hierarchically. These clusters should be compared with each other and then they should be validated. /ifferent clustering methods result in different results. 1o it is necessary to have enough understanding of the data to differentiate between the real clusters and the clusters imposed on the data by the method. <igure #(.% depicts a cluster analysis of customers based on their age and income. ;et us say that this analysis was done for a real estate product. Aroup E is a group with high income old people and can be targeted with a luxurious and restful real estate product. 7n the other hand, cluster / represents market segment that needs a low priced, functional product without any frills.

#(+

CHAPTER 1@ ANALYSIS OF VARIANCE /ANOVA0 1@.0 INTRODUCTION

6nalysis of Bariance (6?7B6) is a method for testing the e uality of two or more population means. The methods for testing hypothesis have already been discussedJ however, with these methods one can compare only two population means. 6?7B6, on the other hand, can be used when more than two means are involved. $t thus helps the researcher to test for the significance of the differences among more than two sample means. $t is also useful in situations where the earnings of different companies have to be compared or the efficiency of different brands of a product have to be compared. $n such cases, more than two samples are compared for analysis. 6nalysis of variance can be unvariate (one!way) or multivariate. Eoth these methods are commonly used in the analysis of experiments. 3e shall discuss these in detail in this chapter. 1@.1 UNIVARIATE ANOVA

;et us use a hypothetical example to explain the concept of 6?7B6. ,<airever. company has come up with three different variants in the package design of its fairness cream product. $t has tested these product package variants in the markets and obtained the sales figures as given in Table #*.#. $n order to use analysis of variance we have to make three assumptions. Fach of the samples is drawn from a normal population. $f the sample si)e is large enough, i.e. when n is greater than &=, we do not need this assumption. The variance of these populations is the same and is represented by d%. Treatments are assigned at random to the test units. This assumption is often overlooked by using pseudo treatments such as occupation, stage of life cycle, and urban or rural residence. These non!randomly assigned treatments greatly increase the possibility of interference by other unaccounted variables associated with the pseudo treatments. These effects will then be wrongly attributed to the pseudo treatments.

T(& M&"(#% 6s we have already seen, in an experiment, we have different groups subjected to different treatments. The greater the effect of treatments on these groups, the greater is the variation between the groups5 means. Eut at the same time, the variation within the group should not change and should be e ual to the population variance. 6?7B6 measures the change in variance between the groups as compared to the variance within the groups. The steps followed in 6?7B6 testing are given below2

#(:

#. %. &.

/etermine one estimate of the population variance from the variance among the sample means. /etermine a second estimate of the population variance from the variance within the samples. @ompare these two estimates, if they are approximately e ual in value, accept the null hypothesis.

Eetween the group variance P -ean 1 uare Treatment 1um of s uare among groups P -1T P !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! /egrees of freedom 1um of 1 uared deviation of group sample means (Rj), from overall sample mean (R) weighted by sample si)e P !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ?umber of Aroups (A)!# P 11T !!!!!! d< (%(,%() /egrees of <reedom

((,() /egrees of <reedom (%,#) /egrees of <reedom

Table #*.# 2 < /istribution

k

` nJ(xj Q xi)%
jP#

-1T P

k!#

3here nj P 1i)e of jth sample lj P -ean of jth sample l P 7verall sample mean #(D

G P number of samples 3ithin the group variance P -ean 1 uare Frror 1um of s uares within groups -1F P !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! /egrees of freedom P 11F !!!!!! d< 1um of 1 uared deviation of each group sample (xij) from the mean of the observations for that group sample (Ri), number of all group sample !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 1um of the sample si)es for all groups minus the number of groups ` ` (xij Q xj)%
jP# iP#

-1T P `
jP#

nj ! k

3here nj P 1i)e of jth sample lj P -ean of jth sample lij P Fach observation in jth group G P number of samples To compare these two variances we can obtain a ratio of them. This ratio is known as < statistic2 <P -1T !!!!!! -1F

$f the null hypothesis is true, then the < statistic should be one, since there is no effect due to the treatments and the between the group variance is e ual to the population variance. $f the null hypothesis is not true, the < statistic tends to be larger than one. The null hypothesis is rejected when < is 4significantly5 larger than one. Table #*.% gives the result of the calculations made for finding < in the ,<airever. problem. S!,&s i$ "(#+s!$%s /80 P #%+)" A P #%+)" B %: %+ &= %D #*=

P& i#%s /I0 # %

P #%+)" C %& %'

& '

&& &'

&= &%

%: %+

T!-,& 1@.1; S!,&s #. F!i &'& p #%+)" 5i"( "( && %i..& &$" p #%+)" p!)4!1&s < statistic (< ratio) has a several curves of distribution. $t is actually part of a whole family of distributions. Fach of these distributions is identified by a pair of degrees of freedom2 the degrees of freedom in the numerator of the < ratio and the degrees of freedom in the denominator of the < ratio. $n figure #*.# a few of the < distributions are illustrated. The degree of freedom for the numerator and the denominator are as follows2 /egrees of freedom for the numerator of the < ratio P number of samples Q # /egrees of freedom in the denominator of the < ratio
k

P ` (nj Q #) P rt Q G
iP#

6s can be seen in figure #*.# , the < distribution is skewed towards the right and becomes more symmetrical as the numbers of degrees of freedom in the numerator and denominator increase. $n 4<airever5 example, let us assume that we re uire a =.=# level of significance. The degree of freedom is calculated as follows2 S#+ )& Eetween Aroups 3ithin groups Total %F % D ## S+2 #. S*+! &s *D.+( (%.+( #%%.( M&!$ S*+! &% &'.+( (.:* '=.*# F !"i# (.D& F p #-!-i,i"3 :.=%

T!-,& 1@.2; T(& #+"p+" #. ANOVA i$ "(& F!i &'& p #-,&2

?umerator has G!# P (&!#)P% degrees of freedom /enominator has ?j Q k P #%!&PD degrees of freedom

#*#

<rom the < tables in the appendix, under the re uired degrees of freedom for =.=# level of significance, we obtain the value of :.=% for <. 1ince the < ratio we have calculated is less than his value, the null hypothesis is acceptable. Thus there is no effect on the sales due to the change in the package design of the 4<airever5 product. 1@.2 MULTIVARIATE ANOVA

The basic approach and logic for multivariate approach are the same as for the univariate approach. -ultivariate 6?7B6 consists of two basic techni ues2 6?7B6 with interaction and 6?7B6 without interaction. $n 6?7B6 without interaction, the effects of the independent variables are independent of each other. $n 6?7B6 with interaction, the effect of the independent variables is dependent on each other. This means that the effect of the independent variables taken together is different from the sum of the individual effect. $n the following section we discuss three experimental designs. Randomi)ed Elock /esigns and ;atin 1 uare methods do not consider interaction effects while the <actorial /esign takes the interaction effect into account. 1@.3 ANOVA FOR RANDOMIEED BLOCLS DESIGNS /RBD0

;et us take the experiment described in the chapter ,Fxperimentation. for understanding randomi)ed block design. To determine the effect of repackaging on 4?imba5 toothpaste sales, we conducted an experiment in which we used the class of city as the blocking factor. ?ow assume that the effect of blocking factors and treatment variables do not interact with each other. $n RE/, we have to calculate between the block variance along with between the treatment variance. -oreover the effects of blocks are taken into consideration for calculating the mean 1 uare Frror. The re uired formulae are as follows2 Eetween the treatment variance -ean 1 uare Treatment (-1T) P 1um of 1 uares Treatments (11T) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<
c

`nj (xj Q xt)%

jP#

!!!!!!!!!!!!!!!!! c!# #*%

Eetween the block variance -ean 1 uare Elocking (-1E) P

r

1um of Frrors (11F) !!!!!!!!!!!!!!!!!!!!!!!!!!!!! d< `ni (xi Q xt)%

iP#

!!!!!!!!!!!!!!!!! r!#

-ean 1 uare Frror (-1F) P

r

1um of Frrors (11F) !!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<

c nis kP#

` ` ` (xijk Q xi!xjexj)%
iP# jP#

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! nr Q r ! c!#

we also have another formula for total means uare (T-1) which we shall not sue at present. Total -ean 1 uare (T-1) P Total 1um of 1 uares (T11) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d< -1E !!!!!! -1F
c nij kP#

<Elocks P

` ` ` (xijk Q xt)%
iP# jP#

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! nr Q# #*&

where nj P 1ample si)e of treatment group j ni P 1ample si)e of block group i xj P -ean of block j xt P Total or ground mean nijP ?umber of respondent (observations) receiving treatment $ and blocking variable j xijkP the Gth observation in treatment $ and block j. nr P total number of observations xr P xj nj 1@.4 ANOVA FOR LATIN SCUARE DESIGN

$n ;atin 1 uare /esign we have two blocking variables and a controlling variable. $n the chapter ,Fxperimentation. we also designed a ;atin s uare design for analy)ing the effect of repackaging on ?imba. The si)e of the city and the si)e of the shop are the two blocking variables. The formulae re uired are -ean 1 uare Row Elock (-1R) P
r

1um of 1 uares Row Elock (11R) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<

r ` (xi Q xt)%
iP#

!!!!!!!!!!!!!!!!! r!#

-ean 1 uare @olumn Elock (-1@) P 1um of 1 uares @olumn Elock (11@) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<
c

c ` (xj Q xt)%
jP#

!!!!!!!!!!!!!!!!! c!# #*'

-ean 1 uare Treatment (-1T) P

t

1um of 1 uares @olumn Elock (11@) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<

t ` (xk Q xt)%
kP#

!!!!!!!!!!!!!!!!! t!#

-ean 1 uare Frror (-1F) P 1um of Frrors (11F) !!!!!!!!!!!!!!!!!!!!!!!!!!!! d< TT1 Q 11R Q 11@ Q 11T !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! (r Q #) (c Q %)

Total -ean 1 uare (T-1) P

r

Total sum of 1 uares (T11) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<

c

` ` (xij Q xtj)%
iP# jP#

!!!!!!!!!!!!!!!!!!!!!! rc!#

-1R <Row P !!!!!! -1F

<@olumns P

-1@ !!!!!! -1F -1T !!!!!! -1F #*(

<Treatment P

where nj P 1ample si)e of row block group j ni P 1ample si)e of column block group i nk P 1ample si)e of treatment group k xj P -ean of row i xj P -ean of column j xk P -ean of treatment group k xt P Total or ground mean 1@.6 ANOVA AITH INTERACTION; FACTORIAL DESIGN

3hen there is reason to suspect interaction between the variables, we have to sue the factorial design method. $n the chapter on ,Fxperimentation. we have discussed the ?imba example again with two P T11 Q 11E Q 11T !!!!!!!!!!!!!!!!!!!!!!!! nr Q r Q c! #

$ndependent variables Q si)e and package design. 3e suspect an interaction between these factors. -ean 1 uare Treatment (-1T) P 1um of s uares treatments (11T) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<
c jP#

`nj (xj Q xt)% P !!!!!!!!!!!!!!!!! c!#

-ean 1 uare Elocking (-1E) P

r

1um of Frrors (11E) !!!!!!!!!!!!!!!!!!!!!!!!!!! d< r `ni (xi Q xt)%

iP#

!!!!!!!!!!!!!!!!! r!#

Total -ean 1 uare (T-1) #**

P
r

Total 1um of 1 uare (T11) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! d<

c nij kP#

` ` ` (xijk Q xt)%
iP# jP#

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! nr Q#

-ean 1 uare Frror (-1F) P 1um of Frrors (11F) !!!!!!!!!!!!!!!!!!!!!!!!!!! d< T11 Q 11T# Q 11T% Q 11F !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! nr Q rc
r c nij kP#

` ` ` (xijk Q xt)%
iP# jP#

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! nr Q# -1T# !!!!!!! -1F -1T% !!!!!!! -1F -1$ !!!!!! -1F

<Treatment# P

<Treatment% P

<$nteraction P 1@.@

ANALYSIS OF COVARIANCE

#*+

7ften a researcher can introduce one or more variables statically into an experiment for the sole purpose of control. There are not real variablesJ they are purely imaginary variables. 3hen a researcher would like to compare two or more groups on a dependent variable controlling one or more relevant variables, 6?@7B6 is useful. Eut it is often used (misused) in uasi!experimental studies or field studies. $n the field, groups often intact and the individuals cannot be randomly assigned between the groups. This means that there are differences in the groups statistically. To make the groups statistically e ual, 6?@7B6 is used to adjust the individual differences. The mathematical methodology for implementing 6?@7B6 is too complicated to be explained here. $t is a linear procedure that should be used with strict adherence to the assumptions. 1o we shall explain its use with an example. 6 study was done where the sample of corporations was classified into three control configurations, based on the distribution of stock ownership. These control configuration groups were <amily $nfluence, management @ontrol and $ndirect <amily $nfluence. 7ne of the objectives of the studies was to assess the effects of managerial power in terms of control configuration groups on managerial tenure and longevity. The researcher felt that the corporate performance of each group of companies might bias the results of the analysis. 1o the researcher used corporate performance as a covariate in 6?@7B6. Thus the characteristics of the groups were essentially e uated. (The use of computer software makes such calculations easy).

#*: