LORD VENKATESHWARA ENGINEERING COLLEGE PULIAMBAKKAM, KANCHIPURAM DISTRICT – 615 605 DEPARTMENT OF MANAGEMENT STUDIES COURSE PORTFOLIO Faculty: Prof.S.Rm.Sokkalingam Subject Code:BA1657 BBBBBBBBBBBBBBBBB Title: Research Methods in Business BBBBBBBBBBBBBBBBB BBBBBA1723 M.B.A – II SEMESTER 1 INDEX Lecture Plan Time Table of faculty Lecture notes Assignment-copies CIA test- 1& answer Scheme CIA test 2 & answer scheme Model test & answer scheme Mark –lists of tests Details of coaching classes Any other details Log-book 2 BA1657 RESEARCH METHODS IN BUSINESS 3 0 0 100 1. INTRODUCTION TO RESEARCH: 8 The hallmarks of scientific research – the building blocks of science in research – the research process for applied and basic research – the need for theoretical frame work – hypothesis development – hypothesis testing with quantitative data. The research design. The purpose of the study: Exploratory, Descriptive, Hypothesis testing (Analytical and Predictive) – cross sectional and longitudinal studies. 2. EXPERIMENTAL DESIGN: 7 The laboratory and the field experiment – internal and external validity – factors affecting internal validity. Measurement of variables – scales and measurement of variables – development scales - rating scale and concept in scales being developed. Stability measures. 3. DATA COLLECTION METHOD: 10 Interviewing, questionnaires etc. Secondary sources of data collection. Guidelines for questionnaire design – electronic questionnaire design and surveys. Special data source: Focus groups, Static and dynamic data-collection methods and when to use each. Sampling techniques and confidence in determining sample size. Hypothesis testing determination of optimal sample size. 4. A REFRESHER ON SOME MULTIVARIATE STATISTICAL TECHNIQUES: 15 Factor analysis – cluster analysis – discriminant analysis –multiple regression & Correlation – canonical correlation – application of SPSS package. 5. THE RESEARCH REPORT: 5 The purpose of the written report – concept of audience – Basics of written reports. The integral parts of a report – the title of a report. The table of content, the synopsis, the introductory section, method of sections of a report, result section – discussion section – recommendation and implementation section. TOTAL : 45 3 TEXT BOOKS: 1. Donald R.Cooper and Ramcis S.Schindler, Business Research Methods, Tata McGraw Hill Publishing CompanyLimited, New Delhi, 2000. 2. C.R.Kothari Research Methodology, Wishva Prakashan, New Delhi, 2001. REFERENCES: 1. Uma Sekaran, Research Methods for Business, John Wiley and Sons Inc., New York, 2000. 2. Donald H.Mc.Burney, Research Methods, Thomson Asia Pvt. ltd. Singapore 2002. 3. G.W.Ticehurst and A.J.Veal, Business Research Methods, Longman, 1999. 4. Ranjit Kumar, Research Methodology, Sage Publication, London, New Delhi, 1999. 5. Raymond-Alain Thie’tart, ET, al., doing management research, sage publication, London, 1999. 4 LORD VENKATESHWARAA ENGINEERING COLLEGE FACULTY : Mr. S.Rm. Sokkalingam Title : Research Methods in LESSON PLAN Business Sub Code : BA 1657 UNIT NO : I 1 The hallmarks of scientific research – the building blocks of science in research – The research process for applied and basic research – the need for theoretical frame work Hypothesis development – hypothesis testing with quantitative data. The research design. The purpose of the study: Exploratory, Descriptive, Hypothesis testing (Analytical and Predictive) Cross sectional and longitudinal studies. 4 5 Page :1 OBJECTIVES: To understand research process, hypothesis and development hypothesis testing. SUBJECT TOPIC 3 : 04.02.2009 TITLE : Introduction to research S.NO 2 Date PERIODS 1 REFERENCE MATERIAL R1 – p 22 2 R1 – p 86 2 R1 – p 103 2 R1 – p 117 1 R1 – 135 5 LORD VENKATESHWARAA ENGINEERING COLLEGE FACULTY : Mr. S.Rm. Sokkalingam Title : Research Methods in LESSON PLAN Business Sub Code : BA 1657 UNIT NO: II Date : 23.02.2009 Page :2 TITLE: Experimental design: OBJECTIVES : To understand experiments, validity factors, of scales and development stability measures. S.NO SUBJECT TOPIC 6 The laboratory and the field experiment Internal and external validity – factors affecting internal validity. 1 REFERENCE MATERIAL R1 – p 144 1 R1 – p 149 Measurement of variables – scales and measurement of variables – Development scales - rating scale and concept in scales being developed. Stability measures. 2 R1 – p 174 2 R1 – p 196 1 R1 – p202 7 8 9 10 PERIODS 6 LORD VENKATESHWARAA ENGINEERING COLLEGE FACULTY : Mr. S.Rm. Sokkalingam Title : Research Methods in LESSON PLAN Business Sub Code : BA 1657 UNIT NO: III 11 Interviewing, questionnaires etc. Secondary sources of data collection. Guidelines for questionnaire design – electronic questionnaire design and surveys. Special data source: Focus groups, Static and dynamic data-collection methods and when to use each. Sampling techniques and confidence in determining sample size. Hypothesis testing determination of optimal sample size. 14 15 Page :3 OBJECTIVES : To understand the techniques of interviewing, designing questionnaire, collection of data and optimal sampling size. SUBJECT TOPIC 13 : 19.03.2009 TITLE: Data collection method S.NO 12 Date PERIODS 2 REFERENCE MATERIAL R1 – p 219, 225 2 R1 – p 245 2 R1 – p 250 2 R1 – p 265 2 R1 – p 286, 292 7 LORD VENKATESHWARAA ENGINEERING COLLEGE FACULTY : Mr. S.Rm. Sokkalingam Title : Research Methods in LESSON PLAN Business Sub Code : BA 1657 UNIT NO: IV Date : 14.04.2009 Page :4 TITLE: A refresher on some multivariate statistical techniques: OBJECTIVES : To understand statistical techniques to test the hypothesis. S.NO SUBJECT TOPIC PERIODS 16 17 18 19 Factor analysis Cluster analysis Discriminant analysis Multiple regression & Correlation 2 2 3 3 20 21 Canonical correlation Application of SPSS package. 3 2 REFERENCE MATERIAL T2 – p 323, T1 T2 – p 337, T1 T2 – p 319, T1 T2 – p 331, T1 R1 – p 403,T1 T2 – p 331, T1 R1 – p 322 8 LORD VENKATESHWARAA ENGINEERING COLLEGE FACULTY : Mr. S.Rm. Sokkalingam Title : Research Methods in LESSON PLAN Business Sub Code : BA 1657 UNIT NO: V Page :5 OBJECTIVES: To understand writing of reports and give recommendations for implementation. SUBJECT TOPIC 22 The purpose of the written report – concept of audience – Basics of written reports. The integral parts of a report – the title of a report. The table of content, the synopsis, the introductory section, method of sections of a report, result section Discussion section Recommendation and implementation section. 24 25 : 04.05.2009 TITLE: The research report: S.NO 23 Date PERIODS 1 REFERENCE MATERIAL R1 – p 346 2 R1 – p 347 1 1 R1 – p 352 R1 – p 371 9 SUGGESTED REFERENCES (Name of the book, name of author(s), and name of publisher and year of publication) T1 Business Research Methods, Donald R.Cooper and Ramcis S.Schindler, Tata McGraw Hill Publishing Company Limited, New Delhi, 2000. T2 Research Methodology, C.R.Kothari, Wishva Prakashan, New Delhi, 2001. R1 Research Methods for Business, Uma Sekaran, John Wiley and Sons Inc., New York, 2000. Type of teaching aids OHP LCD Website Serial numbers 1 to 25 10 TIME TABLE OF THE FACULTY 1. Research methods 4 periods per week 2. Project guidance 4 periods per week 11 Introduction to Research Managerial decisions based on the scientific research tend to be effective. Research is defined as ‘an organized, systematic, data-based, critical, objective, scientific inquiry into specific problem that needs a solution’ Scientific research focuses on solving problems and follows a step-by-step logical, organized, and rigorous method to identify the problems, gather data, analyze them, and draw sound conclusions. The scientific decision is not based on hunches, experience, and intuition. Researchers draw comparable findings with accuracy and confidence for similar issues while analyzing the data. Scientific investigations are more objective than subjective. It helps the managers to identify the most important factors that need specific attention to solve the problem. Scientific investigation and managerial decision making are integral part of decision making. The term scientific research refers to both basic and applied research. Applied research is organized and it is a systematic process to carefully identify the problems, gathering data scientifically and analyzed and conclusions are arrived for effective problem solving in an objective manner. Simple problems do not attract extensive research. Past experiences could offer appropriate solution. The probability of making wrong decision shall be high in the absence of funds required to support good research, insufficient information, and decisions taken on hunches. --- 12 The Hallmarks of Scientific Research The hallmarks of scientific research are listed as follows: 1. 2. 3. 4. 5. 6. 7. 8. Purposiveness Rigor Testability Replicability Precision and Confidence Objectivity Generalisability Parsimony A manager is interested to find out how commitment of his employees can be increased. We shall now see how he applies the eight hallmarks scientifically. Purposiveness: Increase in employee commitment will reduce employee turnover, less absenteeism, and may be increased performance levels, which will benefit the organization. Thus the manager has started the research with a definite purpose. Rigor: Rigorous research requires a good theoretical foundation and a carefully designed procedure. For example, the manager asks his 10 out of 12 employees to mention what would make them to increase their involvement. Based on the response from the employees if the manager takes decision, how employees’ commitment can be increased, then it would not be scientific as detailed below: 1. Few employees’ opinion will not be representing all the employees. 2. Framing and addressing the questions will be biased or incorrect response. 3. Small sample of respondents will reduce the important influence on the organizational commitment. The researcher should collect the right kind of data from an appropriate sample, with less amount of bias to analyze the data so collected. Testability; After collecting the data how to increase the commitment of the employees the researcher develops certain hypothesis, how employee commitment can be increased. He would apply few statistical tests to the data gathered for this purpose. The researcher may have developed a hypothesis to pursue larger opportunities to participate in the decision making to have higher commitment. This hypothesis can be tested using chi-square test, the t-test to be discussed later. Testing logically developed hypothesis to ensure whether the data collected support the educated conjectures (an opinion based on incomplete information) or hypotheses. Replicability: The researcher concludes that participating in decision making is an important factor to influence the commitment of the employee to the organization. When the same type of research is repeated, the result of the hypotheses tested would be the same in similar circumstances. 13 Precision and Confidence: The researcher would like to design the research so that the findings shall be close to reality, as the samples collected will not reflect the entire characteristics. Precision refers to the closeness of findings to ‘reality’ based on the sample. You may call the term confidence interval in statistics. If estimated production days of loss are between 30 & 40 and if the actual is 35, the precision is very close. Confidence refers to the probability that our estimation is correct. E.g. we can confidently claim that 95% of the time our results are true and there would be a chance of 5% chance of our being wrong. This is known as confidence level. The greater the precision and confidence, the investigation shall be scientific and useful. Objectivity: The conclusion drawn should be objective. If the hypotheses of greater participation in decision making will improve the employee commitment, and the result is different and the researcher continue to argue that increased participation would still help, it does not make any sense. Considerable time, energy and money are wasted if findings are otherwise. When the data is objectively interpreted, the research investigation will be more scientific. Generalisability: Generalisability means the applicability of the research findings in one organizational setting to the other. More elaborate sampling design will increase the generalisability of the research, but it will increase the cost of the research. Much applied research pertains to the given organization and the result can be general sable only in identical situations and settings. Parsimony: Simplicity in explaining the problem against complex research frame works. For example, if two or three variables are changed, the commitment of the employees will be increased by 45%, and it would be valuable. If 10 variables have to be changed to achieve 48%, such number is unmanageable and it would beyond the control of the manager to change. The reason for following scientific research method will be less prone to errors compared to simply collecting the data and analyzing them. Obstacles to conduct Scientific Research: In the management and behavioral areas 100% scientific investigations are not possible and the result will not be error free because difficulty may arise in: 1. the measurement and collection of data in the subjective areas viz. feelings, emotions, attitudes, and perceptions 2. quantifying human behavior 3. obtaining representative samples Comparability, consistency and wide generalisability are often difficult to obtain in research. However the research is designed to ensure all the hallmarks of scientific research. We shall study various other limitations later. --14 The Building Blocks of Science in Research The hypothetico-deductive method is one of the primary methods of scientific investigation. The deductive and inductive process of deduction in research is described below: Deduction and Induction: We can find the answers to the scientific investigation either by the process of deduction or the process of induction or by the combination of the two. Deduction is the process to arrive at a reasoned (cause or explanation) conclusion by logical generation of a known fact. e.g. It is a fact that all high performers are highly competent in their jobs. If Gabriel is a high performer, we conclude that he is competent in his job Induction is a process to observe certain facts and conclusions are arrived on that basis. e.g. We know that production process is the main activity of the factories. We then, conclude that factories exist for the purpose of production. The method of starting with a theoretical framework, formulating hypotheses, and logically deducing (logically arrive at by reasoning) from the study is known as the hypothetico-deductive method. e.g. The individual performance of solving puzzles can be increased, if the excessive noise in the environment can be controlled. The investigator begins with this theory. He develops a hypothesis accordingly. The building blocks of scientific inquiry can be seen in fig. 2.1 attached. It includes the process of initially observing phenomena (facts), identifying the problem, constructing a theory as to what might be happening, developing hypotheses, determining aspects of research design, collecting the data, analyzing the data and interpreting the skills. The Hypothetico-Deductive Method: The Seven-Step process in the Hypothetico-Deductive Method: The following seven steps emerge from the building blocks discussed as above: 1. 2. 3. 4. 5. 6. 7. Observation Preliminary information gathering Theory formulation Hypothesizing Further Scientific data collection Data analysis Deduction --- 15 The Research Process for the Applied and Basic Research We need to identify the variables tend to be problematic to the manager. We need to discuss the important aspects viz. – the process of developing the conceptual framework and the hypotheses for testing; and the design involve the following: the planning of the actual study dealing with the aspects as the location of the study the selection of the sample and collection & analysis of the data Please refer to fig.4.1 – The research process for basic and applied research in the annexure. We discuss shall discuss the research process as a step-by-step process, but it is not actually in practice. We conduct the literature search and interviews before formulating the theoretical framework. However we conduct more interviews and / or seek additional information from the literature for clear understanding in order to refine the theory. We shall discuss the steps 1 to 3 of figure 4.1 as follows: 1. the identification of the broad problem area; 2. preliminary information gathering (unstructured and structured interviews) and literature survey; and problem definition Broad problem area: The broad problem area covers the entire situation for the purpose of solving the problem. We may not identify the specific issues at this stage. The specific issues shall be: 1. 2. 3. 4. current problems to be solved areas needed to be improved conceptual or theoretical issues needed to be tightened-up some basic questions, the researcher would like to empirically (based on observation or experience rather than theory) answered e.g. the issue of sexual harassment Example of problem existing currently: Receiving complaints from women for ‘not being treated right’. As the problem is related to gender, the manager would not be able to pin point what the actual problem is. He has to make further identification to identify and solve the problem. Example of a situation requiring improvement: Suppose, the company has already formulated policies on discrimination and sexual harassment. If the company continues to receive legitimate complaints on discrimination, the policies framed could be ambiguous (in understanding or in enforcement) and should be re-framed. Conceptual issues that need to be tightened: The basic researcher has to study sexual harassment to define the concept in definite terms. ‘Any unwelcome sexual advances, requests for sexual favors, and other verbal and physical conduct of a sexual nature’ can be the definition to start with. Answer empirically: When the issue of sexual harassment either perceived or actual and its impact on the consequences for the individuals (e.g. psychological stress) and organizations 16 (poor performance) is explored by gathering the data and testing the relationships. Here we seek some specific answers in respect of the research. The following are some of the examples of broad problem areas: 1. 2. 3. 4. 5. Training programs are perhaps not as effective as anticipated. The sales volume of a product is not picking up. Minority group members in organizations are not advancing in their careers. The daily balancing of accounting ledgers is becoming a continuing concern. The managers for whom it was primarily designed are not using the newly installed information system. 6. The introduction of flexible work hours has created more problems than it has solved in many companies. 7. The anticipated results of a recent merger have not been forthcoming. 8. Inventory control is not effective. 9. The installation of a MIS keeps getting stalled. 10. The management of a complex, multi-departmental team project is getting out of hand in the R & D department of the company. Once the researcher collects the preliminary data, the broad problem area will be narrowed down to specific problems. This can be done through interviews and literature research. Preliminary Data Collection We know that the structured / unstructured interviews and library research would help the researcher to define the problem more precisely, evolve a theory by removing the possible variables that are not relevant to the problem. The kind of information required by the researcher shall be classified into three types as follows: 1. Background information of the organization – the contextual factors. 2. Managerial philosophy, company policies, and other structural aspects. 3. Perceptions, attitudes, and behavioral responses of organizational members and client systems (as applicable). Background details can be obtained from the company’s published records. The web site of the company, its archives and other sources. The company policies, procedures, and rules can be obtained from the organization’s records and documents – secondary data. The information viz. perceptions and attitudes of employees can be best obtained by talking to them. Collecting such data for research from the actual place of occurrence of events are called - primary data. Background information on the Organization: The researcher could be from an outside agency. He should be well acquainted with the background of the company before conducting the first interview of the staff concerned. Background information can be obtained from trade publications, the Census of Business and Industry, Directory of Corporations, several other business guides and services, web etc. The background information shall include: 1. The origin and history of the company. 17 2. 3. 4. 5. 6. 7. Size in terms of employees, assets or both. Charter – purpose and ideology. Location – regional, national, or other. Resources – human and others. Interdependent relationships with other institutions and the external environment. Financial position during the previous 5 to 10 years and financial data. The researcher can knowledgeably interact with the personnel in the company during the interview and can raise appropriate issues related to the problem. The researcher is aware that the problems faced by a given industry are not unique. Others also face similar problems viz. competition from the foreign producers, consumer resistance to spending money etc. The background information thus, will help the researcher to focus more questions towards strategies (such as sales and advertising efforts) developed by the company to promote sales in the face of foreign competition. Information on Structural Factors and management Philosophy: The researcher can directly ask questions to the management in respect of information on company policies, workflow, management philosophy etc. The responses could be conflicting and contradictory. Frequent contradictions may indicate poor communication or misinterpretations by the members of the organization’s philosophy, goals, values and so forth. It will give an idea to the researcher upped what extent the difference in perception exist in the organization. Such information gathering will be useful when newly installed system and procedure do not produce the desired results. The failure of many new technologies, well meant benefit policies, strategic plans, or marketing or production practices is often due to misunderstanding or misperceptions against the desired goals and motives of the top management of the organization. Once the misperceptions are cleared, the problem will automatically vanish. The researcher will come to understand the priorities and values of the company by questioning the managerial and company philosophy. For example: either product quality is really deemed important to the company or only a lip service. Whether the company has a short term or long term goals Whether controls are so tight that creativity is prevented so loose or conducive to good performance. Whether the company always wants to play it safe to take calculated risk; and whether it is people oriented or solely profits-oriented. Often, certain parts of a structure also influence the problem. Some of the structural factors are cited below: 1. 2. 3. 4. 5. 6. 7. Roles and positions in the organization and number of employees at each job level Extent of specialization Communication channels Control systems Coordination and span of control Reward systems Workflow systems and the like 18 The respondents’ perceptions or understanding the structural variables may not conform to the formal written structural policies or procedures of the company. But it can be appropriate leads for interviewing various employees at various job levels in the company. Perceptions, Attitudes, and Behavioral Responses: We can get the response from the employees about their perceptions of the work and work environment and their attitudinal and behavioral responses through appropriate questionnaires. The researcher can use the structured or unstructured interviews with various levels of employees in the organization. The problem definitions can be arrived accurately, provided we have an understanding on the attitudinal and behavioral reactions of the employees. Attitudinal factors refer to people’s beliefs about and reactions to the following: 1. 2. 3. 4. 5. 6. 7. 8. 9. Nature of the work Workflow interdependencies Superiors in the organization Participation in decision making Client systems Co-workers Rewards provided by the company Opportunities for advancement in the organization Organization attitudes toward employee’s family responsibilities , company’s involvement with community, civic and other social groups 10. Company’s tolerance of employees’ taking time off from the job. Behavioral factors include actual work habits such as industriousness, extent of absenteeism, performance of the job etc. The researcher should encourage the respondents at the interviewing stage to talk about their jobs, other work and non-work-related factors and their attitudes values, perceptions, and behaviors, because these factors may influence problem area. The interviewer will quickly become familiar with the environment by talking to the people in the company. The researcher has to deal with certain aspects only in depth, depending upon the type of problem investigated, and the nature of initial responses received. For example, if the problem as identified by the manager is related to ‘individual’s attitudes and behaviors’, then the value system, the corporate culture, and the employee’s perception may have to thoroughly research than the structural aspects. The important reason to gather information on values, structures and processors is to identify the root of the real problem. For example, a company is introducing employee stock ownership plans (ESOP). All the employees are not interested in this program. Instead of making the package more attractive, we need to talk to the employees who may reveal that they perceived the ESOP merely as a tool to postpone the takeovers & save taxes and no true opportunities are provided to the employee involvement and participation. The manager is able to understand the real issues (common fear 19 of the employees in this case) instead of solving the issue on the surface symptoms (making cosmic changes in the package to make it more attractive). Example: A manager may think that the inventory and production cost can be lowered if the just-in-time (JIT) system is refined whereas the type of machinery used in the production process could be the main reason. When the interviews are over, the researcher has to tabulate the data collected and conclude whether there are any recognizable patterns in the response. Sometimes inadequate tools, insufficient lighting, untrained personnel etc., have been firmly expressed by the many workers, the researcher will survey the literatures (secondary data) to how other researchers perceived such factors in other work situations and defined the problem before arriving at solutions. Literature Survey: Literature survey is the documentation of a comprehensive review of the published and unpublished work from secondary sources of data in the areas of specific interest to the researcher. Library is the main place to access the secondary data. Books, journals, newspapers, magazines, government publications, financial, marketing and other reports are also examples for secondary data. The researcher can look into the data simultaneously conducting structured and unstructured interviews. With computerized database, the literature search can be much faster and easier. Reasons for the literature survey: Some variables might not be identified in interviews, but influence the problem critically. The research done without considering the unidentifiable variables is a wasteful exercise. Therefore the researcher has to make a thorough research work relating to the particular problem area. Example: The company designs appropriate tests to assess the applicant analytical skills, judgment leadership, motivation, oral and written communication skills, etc. They are also highly paid. But the employee turnover was very high. We can’t unearth any idea by reviewing the interviews conducted. However we can review the literature might indicate that when the employees have unmet job expectations, they will be inclined to leave the company. Further talking to the officials of the company, it could be found out that realistic job previews were not offered to the applicants at the time of interview. The important factor cited herein might not have been unearthed but for the literature survey. We should go for a good literature survey for quality research to trace the origins and progress of technology and predicting where it is heading for the future. A good literature survey provides the foundation to develop comprehensive theoretical framework from which the hypotheses can be tested. 1. A good literature survey takes care of the following: 2. Important variables likely to influence the problem situation are not left out in the study. A clear idea emerges as to what variables would be most important to consider (parsimony), why they would be considered important, and how they should be investigated to solve the problem. Thus the literature survey helps the development of the theoretical framework and hypotheses for testing. 3. The problem statement can be made with precision and clarity. 20 4. Testability and replicability of the findings of the current research are enhanced. 5. One does not run the risk of running of ‘reinventing the wheel’ i.e. wasting efforts on trying to rediscover something that is already known. 6. The scientific community perceives the problem investigated as relevant and significant. Conducting the literature survey: 1. Identify various published and unpublished materials that are available on the topics of interest. 2. Gather the relevant information either by going through the necessary materials in the library or by getting access to online sources. Identifying relevant sources Published work Computer online system Global business information* Published articles – news papers and periodicals* Conference proceedings* *Available on data bases Computer databases include bibliographies, abstracts, and full texts of articles on various topics (business). Databases are also available to obtain statistics, marketing, finance etc. The forms of databases that are available in the internet are as follows: 1. The bibliographic databases: It contains the name of the author, the title of the article (or book), sources of publication, year, volume and page numbers. 2. The abstract databases: It provides an abstract or summary of articles in addition to the above. 3. The full text databases: It provides the full text of the article. Advantages of online searches: We can save plenty of time; data is comprehensive in listing, and review of references help to focus on materials that are important to research. If sources are not known, the data bases can be found with the search strategies in the internet. Problem definition After conducting the interview, the researcher narrows down the problem and define the issues more clearly. The researcher studies and pin points the critical issue or the problem clearly. He defines the problem where the gap exists between the actual and the desired or ideal state. Some examples of defining the problem are given below: Severe decline in productivity or the company is fast losing its market. Here the goal is to rectify the problem with the sense of heightened urgency. Considerable interest in attracting highly qualified engineers to the firm or enhancing the quality of life for their employees. 21 Researchers do not define the symptoms of the problem. For example: Low productivity could be a symptom. Real problem could be low morale and motivation of employees because they were not recognized as a valuable contributors and do not get appreciation for their good work. Finding right answers to the wrong problem definition will not help. Managers tend to describe the problem in terms of symptoms. But researchers have to identify the problem more accurately. The real issue in the above example is morale and motivation. The consequences of the problem are low productivity. The antecedent (exists before another) of the problem (i.e. contributing factor) is non recognition of employees contribution. Central issue of the problem needs to be addressed. Though more money is given, better machines installed to increase productivity, desired results will not be achieved if the right problem is not addressed. Problem definition or problem statement A clear, precise, and succinct (clearly and briefly) statement of the question or issue have to be investigated with the goal of finding an answer or solution. Problem definitions could pertain to: Applied research: Existing business problems Manager may look for improvement in the existing problem Basic research: Areas where some conceptual clarity is needed for better theory building or Sometimes researcher is trying to answer the questions empirically because of his interest in the topic of research. --- 22 The Need for Theoretical Framework It is a conceptual model. It makes logical sense of the relationships of several important factors, identified as important to the problem. It investigates logical believes with published research duly taking into the boundaries and constraints. Theoretical framework discusses the interrelationship among the variables. From the theoretical framework testable hypotheses can be developed to ascertain whether the theory formulated is valid or not and tested through appropriate statistical analysis. Developing a good theoretical framework is central to examine the problem under investigation. Variables The values can differ at various times for the same object or problem or at the same time for different objects or persons. For example, production units, absenteeism, motivation etc. Types of variables: 1. 2. 3. 4. The dependent variable The independent variable The moderating variable The intervening variable Dependent variable Definition: The variable that is primary interest to the researcher is called as dependent variable. The researcher has to understand and describe the dependent variable, or to explain its variability, or predict it. It is possible to find a solution to the problem by finding what variables influence the dependent variable. The researcher is interested in quantifying and measuring the dependent variable and other variables that influence the dependent variable. Example 1. Introduction of a product after test marketing – here, sales is the dependent variable. 2. Investigating the debt equity ratio of a manufacturing company – ratio of debt to equity is the dependent variable. 3. Employees are not loyal to the organization – variance found in the level of organizational loyalty of employees. We need to find out what variable affect the variance in loyalty. There may be more than one dependent variable, for example: 1. There is always conflict between quality and volume of output. 2. Low cost production and customer satisfaction The manager is interested to know the factors that influence all the dependent variables of interest and how some of them differ in regard to different dependent variables. The researcher uses multivariate analysis in this regard. 23 Independent variable Definition: An independent variable is one that influences the dependent variable either a positive or negative way. Each unit independent variable increase, there is increase or decrease in the independent variable, i.e. the variance in the dependent variable is accounted for by the independent variable. Example for independent variable Diagram - relationship between the independent variable (new product success) and the dependent variable (stock market price). S Stock Market price New product Success Independent variable Dependent variable Diagram of relationships between the independent variable (managerial values) and the dependent variable (power distance). Power distance Managerial values Independent variable Dependent variable Moderating variable Definition: The presence of a third variable (the moderating variable) modifies the original relationship between the independent and dependent variables. The moderating variable has a strong contingent effect on the independent – dependent variable, i.e. the presence of moderating variable modifies the relationship between the independent and dependent variables. No. of No. of rejects rejects Availability of reference manuals Independent variable Dependent variable Interest and Inclination Moderating variable 24 The distinction between an independent variable and a moderating variable are discussed below: At times confusion may arise as to which variable is to be treated as independent variable and when it would become a moderating variable. We can understand the moderating variable from the following situations: Situation 1 A research study indicates that the better quality of the training programme in an organization and the growth needs of the employees, the greater is their willingness to learn new ways of doing things. Following is the diagram of relationship among three variables viz. workforce diversity, organizational effectiveness and managerial expertise. Workforce diversity Workforce Organisational Organisational effectiveness Diversity effectiveness Independent variable Dependent variable Managerial Expertise Moderating variable Situation 2 According to another research study, the willingness of employees to new ways of doing things is not influenced by the quality of training programs offered to all people with out any distinction. Only high growth people will be longing to learn. In the above referred situations we have the same three variables. In the first training program and growth need strength are independent variables that influence the employees’ willingness to learn (dependent variable). 25 In the second case, the growth need strength becomes a moderating variable. When the quality of the program is improved, only high growth need persons show willingness and adoptability. Therefore the relationship between the independent and the dependent variables have now become contingent on the existence of a moderator. Please refer to the figures in the next page for differences in the effect of independent and moderating variables. Intervening variable Definition: The intervening variable surfaces between the times the independent variables start operating to influence the dependent variable and the time their impact is felt on it. There is a temporal quality or time dimension to the intervening variable. The intervening variable surfaces as a function of the independent variable(s) operating in any situation, and helps to conceptualize and explain the influence of the independent variable(s) on the dependent variable. Please see the following example: Diagram of the relationship among the independent and, intervening, and dependent variable: Diagram of relationship among the independent, intervening, moderating and dependent variables: 26 The following five basic features should be incorporated in any theoretical framework: 1. The variables considered relevant to the study should be clearly identified and labeled in the discussions. 2. The discussions should state how two or more variables are related to one another for important relationships that are theorized to exist among the variables. 3. If the nature and direction of relationships can be an indication in the discussion as to whether the relationship should be positive or negative. 4. There should be a clear explanation of why we would expect these relationships to exist. The arguments could be drawn from the previous research findings. 5. A schematic diagram of the theoretical framework should be given so that the reader can see and easily comprehend the theorized relationship Theoretical framework We have studied different kind of variables and relationship among the variables. The theoretical frame work is logically developed, described, and elaborated net work of associations among the variables relevant to the problem situation. We use the processes viz. interviews, observations, and literature survey. Experience and intuition guide to develop the theoretical framework. One has to correctly identify the problem and the variables contributing to it. Then elaborate network of associations among the variables, construct hypotheses and testing. The testing of hypotheses would tell you to what extent the problem can be solved. The literature survey provides a solid foundation for developing the theoretical framework. The theoretical framework elaborates the relationship among the variables. The literature survey identifies important variables as determined by the previous research findings. It provides the nature and direction of relationships. The components of theoretical framework Theoretical framework identifies and labels the important variables. It logically describes the relationships among the variables and elaborating them. Moderating variables should be explained as to what specific relationship they would moderate and an explanation why they operate as moderators. An explanation should be given why an intervening variable is treated as intervening variable. --- 27 Experimental Designs Scenario A CEOs are accountable for performance than getting compensation irrespective of performance. Scenario B A study of absenteeism and steps taken to curb reveals that companies provide incentives to employees. There may be a casual or indirect connection between one or two specific incentives and absenteeism. Scenario C When workers are laid of, there shall be drop in commitment of workers retained. *** We can use experimental designs in researching the aforesaid issues. Experimental designs are set-up to examine possible cause and effect relationships among relationships in contrast to correlation studies. Correlation studies examine the relationships among variables without necessarily trying to establish if one variable causes another. All the following three conditions have to be satisfied to establish that variable ‘x’ causes variable ‘y’. 1. Both ‘x’ and ‘y’ should co-vary 2. ‘x’ should precede ‘y’ and 3. No other factors should possibly cause the change in the independent variable ‘y’. It is necessary to establish casual relationship between two variables. Several variables co-vary with the dependent variable and they have to be controlled. i.e. the variable ‘x’ and the variable ‘x’ alone cause the dependent variable. It is not always possible to control all the covariates while manipulating the casual factor, where events flow normally and naturally. It is possible to isolate (separate) the effects of a variable in a tightly controlled artificial setting or lab setting. Example Suppose a manager believes that if the accounting department is completely managed by chartered accountants (CAs), the productivity can be increased. The manager has an option to recruit CAs and transfer non-CAs to other departments. The implication of this decision shall be: 1. Disrupt the present work 2. More people will have to be trained 3. Work may slow down 4. Non-CA employees may be upset 28 The above hypotheses can be tested in an artificially created settings viz. three groups. 1. Those with CAs 2. Those without CA qualification 3. Mixed group with CAs and non-CAs The result of the tests may indicate that the first group with CAs causes productivity increase. The manager can take a strategy to gradually transfer those without CA qualification and recruit CAs. Experimental Designs Experimental designs fall into two categories. 1. Lab experiment 2. Field experiment Experiments done in an artificial environment is known as lab experiments. Experiments done in natural environment is known as the field experiment. The Lab Experiment The possible effects of other variables on the dependent variable have to be accounted for, to determine the actual effects of the investigated independent variable on the dependent variable. We need to manipulate the independent variable to establish the extent of its casual effects. The In an artificial environment (the laboratory) we can effectively control and manipulate the independent variable. Control Say, factor A, can influence the dependent variable ‘y’ in the relationships between ‘x’ and ‘y’. We will not be able to assess the extent to which ‘y’ occurred because of ‘x’ as we do not know how the total variance was caused by the presence of factor A. Example A bank trains the new recruits to market banking products. Some will function effectively than others because of previous experience. The term ‘experience’ is a contaminating factor. To assess the effectiveness of training, we need to control the contaminating factor by excluding the experienced from the training programme. Manipulations of the independent variable We need to manipulate the independent variable to examine its casual effect on a dependent variable. Manipulation means, creating different levels of the independent variables to asses the impact on the dependent variable. 29 Example ‘Depth of knowledge of various manufacturing technologies is caused by rotating the employees on all the jobs on the production line and in the design department over a four week period’. The effect in depth of knowledge shall be: 1. Rotating one group and expose them in all the systems – greatest gain. 2. Rotating one group partially – some significant increase 3. No rotation – lowest increase Control the contaminating factors and measure the depth of knowledge of the above groups before and after manipulation, to asses the extent to which the treatment caused the effect. Example ‘We want to test the effects of lighting on production levels among 60 sewing machine operators’. Measure the production levels of all the operators for 15 days with usual lighting of 60 watt lamps. Split groups into 20 members each as follows: 1. Allow one group to work under the same condition – 60watts lamps 2. One group with 75 watts lamps and 3. The third group with 100 watts lamps Analyse each group’s production for the next 15 days. If your hypotheses of better lighting increases the production level is correct, then there should be: 1. No increase in production level for the first group 2. Small increase in production for 75 watts lighting and 3. Greater increase in production for 100 watts lighting In the above example, we have manipulated the independent variable ‘lighting’. The manipulations of independent variable are known as treatment and the results of the treatment are called treatment effects. Example ‘Whether paying the piece rate would increase the production rates than the hourly rates? You must make sure that switching over to piece rate system will help achieve enhanced productivity. Test the casual relationship in a lab setting. If results are encouraging, conduct the experiment in field settings. Find out the possible factors (previous experience, gender difference, age etc.) that affect the production levels of workers and try to control them. The cause and effect relationship impacted by the variables can be controlled in two ways: 1. Matching groups 2. Randomization 30 Example We can understand matching group and randomization from the following example: 1. Set four groups of 15 workers each 2. Use the first group as a control group 3. Manipulate the other three groups Matching groups If there are twenty women out of sixty workers, assign five each to four groups. The effect of gender factor is distributed to the four groups. Match age and experience and distribute the workers accordingly. As suspected contaminated variables are matched across the groups, we are confident that variable ‘x’ alone causes variable ‘y’. however, we are not sure whether we have controlled all the contaminated or nuisance factors. We may not even we have an idea about them. Therefore the best way is to randomize. Randomization Draw from the pot and assign the sixty members randomly to the four groups. Assign 15 names to the first group, 15 to the next group and so on or first person to the first group, the second person to the second group and so on. In this process every member will have an equal chance and assignment to any particular group is random. Thus the confounding variables are randomly distributed. The variables of age, sex and previous experience (the controlled variables) will have an equal probability of distribution among the four groups. i.e. each group is comparable to each other. The confounding variable is controlled by distributing across groups along with other unknown factors. We can manipulate the piece rate system as follows: Group Treatment Experimental group 1 Experimental group 2 Experimental group 3 Control group Re.1/= per piece Rs.1.50 per piece Rs.2.00 per piece old hourly rate Treatment effect (% of increase over Pre-piece rate system) 10 15 20 0 The piece rates (the treatment), is the cause for the increase in the number of pieces produced. The cause and effect relationships confounded by other ‘nuisance’ variables are controlled by the process of randomly assigning members to the groups. Hence, we have high internal validity or confidence in the cause and effect relationships. Advantages of randomization Matching: Matched to control the differences deliberately. 31 Randomization: Inequalities are distributed among the groups based on the laws of normal distribution. Hence, we need not have to concern about any known and unknown confounding factors. Disadvantage of randomization Matching some the critical factors to all groups could be difficult while conducting an experiment. When a confounding variable is known for example, there are only two women in a group of four experimental design, we will not be able to match the gender across all the groups. However, randomization takes care to spread all the contaminating factors across all the groups. Contaminating variables are controlled through the process either matching or randomization, and manipulation of the treatment. Internal validity Internal validity refers to the cause and effect relationship. The relation ship shall be casual when there is high internal validity. Casual effect can not be felt where the internal validity is low. In a lab environment, where cause and effect relationships are substantiated, the internal validity can be higher. You may be participating in a lab experiment. You will be told a cover story. i.e. you will be informed the reason for the study and your role in it and not the actual story or true picture. Selection of participant is by randomization. Participants are moved into lab setting, provided with details of study and tasks to perform. Questionnaire or test will be conducted to the participants before and after the task is completed. When the experiment is over, you will be explained about the experiment and all your questions will be answered. The result of the study will indicate the cause and effect relationship between the variables under investigation. External validity or generalisability of lab experiment ‘Whether the result of cause and effect relationships in a lab experiment will hold well in the organizational setting?’ Example In a lab experiment, groups are given production tasks to screw bolts and nuts. Productivity of groups paid with piece rates are better than the groups paid with hourly rates. Can this result hold in the organizational setting? Organizational setting is far more complex. We can’t control several confounding variables viz. experience. The cause and effect relationships found in a lab experiment in such a situation, can’t likely to hold good or true in the field setting. Therefore we conduct field experiment to test the casual relationships. 32 Field experiment Experiment is done in the natural environment. But treatment is given to one or more groups. Even variables can’t be controlled, treatment can be manipulated. Example There are three shifts in a factory. We can find the effect of piece rate system on the dependent variable ‘productivity’ as describe below: I shift - controlled group II shift and III shift – two different treatment or same treatment (different or same piece rate) The cause and relationship found in the above conditions will have wider generalisability under similar production settings. However we are not sure whether the piece rates alone was the cause for the increase in productivity. We may not be sure to what extent the piece rates alone was the cause to increase productivity as other confounding variables could not be controlled. External validity External validity refers to the extent of generalisability of the results of a casual study to other settings, people or events. Internal validity refers to the degree of our confidence in the casual effects, i.e. the variable ‘x’ causes variable ‘y’. The results of field experiments can be generalizable to other similar settings. Field experiments have more external validity but less internal validity. However, we are still not sure to what extent variable ‘x’ alone causes variable ‘y’. In lab experiment we can be sure that variable ‘x’ causes variable ‘y’ because confounding variables are kept under control. We do not know to what extent our lab studies can be generalized i.e. lab findings how validly represent the realities in the outside world. Trade off between internal validity and external validity If you want high internal validity, you should be willing to settle for lower external validity and vice versa. In order to have validity in both internal and external, casual relationships have to be tested in a tightly controlled artificial or lab setting and test the casual relationship in a field experiment once the relationship is established in the lab setting. Gender differences in leadership styles, managerial aptitudes etc., in the management area can not be assessed in a lab experimental designs because the factors in lab setting are not found in field study. Field studies in management area are not taken because of resultant unintended consequences i.e. employees becoming suspicious, rival or jealous due to differences among various departments over a period of time. 33 Factors affecting internal validity Confounding factors still may found in the best designed lab studies and could offer different explanations as to what is causing the dependent variable. The possible confounding variable poses a threat to the internal validity. The major seven threats to internal validity are the effects of: 1. 2. 3. 4. 5. 6. 7. History effects Maturation effects Testing effects Instrumentation effects Selection bias effects Statistical regression Mortality History effects When experiment is in progress certain events may occur. It would confound the cause and effect relationship between the two variables. Thus the internal validity is affected. Example When the lab experiment is in progress, the company may release an advertisement ‘buy one and get one free’. This advertisement may increase the sales substantially. You will not be able to find the cause and relationship to the dependent variable ‘sales’ as per original lab experimental design. Illustration of history effects on cause and effect relationships Maturation effects Cause and effect inferences can also be contaminated by the effect of the passage of time. There could be maturation effect on the dependent variable, purely because of time i.e. growing older, getting fired, feeling hungry and getting bored. Example ‘Increase in the efficiency of workers would result within 3 months of time if advanced technology is introduced in the work setting’. 34 Employees would have gained experience over a period of time which results in better job performance and improved efficiency. Therefore we can’t claim that advanced technology alone increased the efficiency of workers. Illustration of maturation effects on cause and effect relationships Testing effects Pre-tests are conducted to test the effects of treatment i.e. a short questionnaire eliciting the feelings and attitudes. Then treatment is given. Then the post test is given. The difference between the pre-test and the post tests are due to the treatment. Exposure to pre-test influences the responses on the post test and would adversely impact on the internal validity. Example ‘Challenging job is expected to cause increases in job satisfaction in their current job. It will sensitize people to the job satisfaction’ When challenging job is introduced, the employees will respond to the post test with the different frame of reference, had not they been sensitized to the issue of job satisfaction through the pretest. It is called testing effect. Testing effects affect internal validity. Here the pre-test could confound the cause and effect relationship by scrutinizing the respondents to the post test. Selection bias effects Improper or unmatched selection of subjects is a threat to internal validity. Example A group has to work for about two hours in a room with a mild stench (a strong and very unpleasant smell) in an experimental condition. Some workers will be volunteering. An ethical researcher may have to disclose the condition of working with a mild stench to the prospective participants, who are not willing to participate. Some volunteers may be interested to participate 35 for incentives. However the volunteers selected later (new comers) may differ the originally selected and the new comers responses to the treatment may be quite different. Therefore newcomers, volunteers, and others who can’t be matched with control groups would pose a threat to internal validity in certain types of experiment. Statistical regression When the members selected for the experimental group have extreme scores on the dependent variable, the effects of statistical regression is brought. For instance, the researcher should not choose subjects or participants with low (score higher) or extreme high (score lower) abilities for the experiment because both of them will score close to the mean on the post test, i.e. ‘regressing towards the mean’. It will not truly reflect the cause and effect relationship. Mortality Another confounding factor on the cause and effect relationship is mortality or attrition when the experiment is in progress. Attrition means the participant dropping out of the experiment. When the composition changes across the groups, it is difficult to compare between the groups. We will not be able to show how much of the effect observed, arose from the treatment. The members who stayed on the experiment could have reacted differently from those who dropped out. Example A manager has chosen eight trainees each for three different training programmes cited below, to devise effective sales strategies. 1. Field training 2. Indoor 3. Mathematical models and simulation. Three persons dropped out of the training programme from the first group, one from the second group and two from the third group for the reason ill health, family exigencies, transportation problem and a car accident. When people are dropping out of the programme it is impossible to compare the effectiveness of the training programme. Identifying threats to internal validity Example Experimental Design: The democratic style of leadership enhances the morale of the employees well. (Morale: the level of confidence and sprits of a person or a group). Experimental group: 1. Control group – autocratic leader 2. Experimental group – democratic leader 3. Experimental group – Laissez-fair leader 36 You have to administer a pre-test before giving treatment. Control group will have no pre-test and not subjected to treatment. Two members from group 2 exited and expressed ‘great’ and ‘performance bound to be high in this group. Two members each in group 1 and 3 left after first hour and said they had to go urgently and can no longer participate. After two hours post test was administered (conducted) on the same line the pre-test was conducted. History effects It is difficult to separate how much increase in morale was due to participative condition alone and how much to the sudden enthusiasm displayed by the two members in group 2. Maturation effects Maturation will have no effect on morale in this situation as passage of time. Testing Control group was not given the pre-test. It is incorrect to compare the experimental scores with the control group. Instrumentation There is no instrumentation bias because the same questionnaire has measured the morale both before and after the treatment of all the members. Selection bias There is no bias as members have been randomly assigned to all the groups. Statistical regression We can assume that members were selected randomly from a normally distributed population. Hence the issue of statistical regression contaminating the experiment does not arise. Mortality Since members dropped out of two experimental groups, the effects of mortality could affect internal validity. Internal validity in case studies There may be several threats to internal validity in a tightly controlled lab experiment. We can’t draw conclusions from case study that describes events occurred during a particular time. A well designed experimental study can indicate a possible casual relationship provided: Members are randomly assigned to experimental and control groups and Manipulated the treatment successfully It is impossible to say what and which factor causes another in respect of case studies. 37 Example There were several causes attributed to ‘Slice’ the soft drink introduced by Pepsi Co. Inc. This product was not taking off after initial success. Among the reasons given are: 1. Cut advertisement for slice 2. Operating on the mistaken premise that juice content in ‘Slice’ would appeal to health conscious buyers 3. PepsiCo’s attempts to milk the brand quickly 4. Several strategic errors made by PepsiCo 5. Underestimation of the time taken to build the brand and the like The above causes provide the basis for developing a theoretical framework to explain the variance. But we can’t conclude the cause and effect relationships determined from the anecdotal (unproved events). Factors affecting external validity Internal validity raises issues whether the treatment alone or some additional or extraneous factors cause the effect. External validity raises issues about generalisability of the findings to other settings. The extent to which the experimental design differs from the settings to which the findings are to be generalized is directly related to the degree of threat it poses to external validity. Example Subjects in a lab experiment might be given a pre-test and a post-test. Pre-test followed by a post test is rarely administered to the employees. Thus the effect of treatment will not be the same in the field and the external validity suffers a diminution (reduction). Another threat is selection of employees. The type of employees selected for the lab testing would be different from the types of employees recruited by the company. The subject (employees) selection and its interaction with the treatment would also pose a threat to external validity. When the lab experiment condition is close to compatible with the real situation, the external validity shall be high. Example A university may give a task to the students and shall manipulate to study the effects on their performance. The findings from this experiment can not be generalized because; the nature of the job in the field may be quite different from lab setting. Review of factors affecting internal and external validity We have so far studied the seven contaminating factors, viz. history, maturation, testing, instrumentation, selection, statistical regression, and mortality. We can reduce the biases or the effect of the contaminating factors by enhancing the level of sophistication of the experimental 38 design to increase the internal validity of the experimental results, as discussed below. These designs could be expensive and time consuming. Types of experimental designs and internal validity The following are the types of experimental designs to guard against the seven contaminating factors as described above. 1. When the experiment is done in a short span of time, there are fewer chances to affect, history, motorization and mortality effects. 2. When the experiment is lasting only an hour or two, there is no effect on all the seven contaminating factors. 3. When the experiment is spread over a period of time, say several months, more number of confounding factors will be encountered (faced). 4. Quasi-experimental designs Some studies expose an experimental group to treatment and measure its effects. 1. 2. 3. 4. It is the weakness of all designs It does not measure the true cause and effect relationship There is no comparison between groups No recording of dependant variable prior to experimental treatment and how it changed after treatment. There is no scientific value in determining the cause and effect relationships. The following are the quasi-experimental designs: 1. Pre-test and post test experimental group designs 2. Post test only with the experimental group. 3. Pre-test and post test experimental group design The experimental group (without the control group) can be given a pre-test and post test to measure the effects of the treatment. Group Pre-test score Experimental group O1 Treatment X Post test score O2 Treatment effect in respect of the above design is: (O2 – O1). The effect of treatment can be measured by measuring the difference between the post test and the pre test (O1 - O2). Testing and instrumentation effects might contaminate the internal validity. If the experiment is extended over a period of time, history and maturation effects may also confound the results. 39 Post tests only with the experimental groups Experimental groups are exposed to only the post test. Testing effects have been avoided in respect of the control group. We must ensure that the two groups are matched for all possible contaminating ‘nuisance’ variables to determine the true effects of treatment. Randomization would take care of this problem. The following are the possible threats to validity in post tests only: 1. When two groups are not matched or randomly assigned selection biases could contaminate the results i.e. the differential recruitment of persons making up the two groups would confound the results. 2. Mortality (dropouts of individuals) can also confound the results. Group Treatment Experimental group Control group Outcome X no treatment O1 O2 Treatment effect in respect of the above design is: (O1 – O2). True experimental design The experimental and control groups are exposed to pre test and post test. The experimental group alone exposed to treatment. Measuring the difference between the post test and the pre test scores of the two groups will give the net effect of the treatment. Both the groups are exposed to pre test and post test as well randomized. Hence the history, maturation, testing, and instrumentation effects (contaminating factors), have been controlled. We have also controlled the effects of selection biases and statistical regression. Mortality could be a problem for this design where the experiment takes several weeks. Pre test and post test of control groups: Group Pre-test score Experimental group Control group O1 O3 Treatment X no treatment Post test score O2 O4 Treatment effect = [(O2 – O1) - (O4 – O3) Solomon Four-group Design In this experiment two experimental group and two control groups are set. One experimental group and one control group can be given both pre test and post test. The other two groups will be given only the post test. The effect of treatment can be calculated several ways. When we get the same results in each different calculation, we can attribute (a characteristic quality) the 40 effects of the treatment. This design is the most comprehensive with least number of problems in respect of internal validity. Solomon Four-group Design and threats to internal validity Group Experimental 1 Control 1 Experimental 2 Control 2 Pre-test score O1 O3 Treatment Post test score X no treatment X no treatment O2 O4 O5 O6 Treatment effect (E) could be judged by E = (O2 – O1) E = (O2 – O4) E = (O5 – O6) E = (O5– O3) E = [(O2 – O1) - (O4– O3) If all E’s are similar, the cause and effect relationship is highly valid. Now let us examine the threat to the internal validity in the Solomon four-group design: 1. Subject or members have been randomly selected and randomly assigned to groups. This removes statistical regression and selection biases, 2. If score of O3 and O4 remain the same, we can establish neither history, nor testing, nor instrumentation, nor statistical regression, nor mortality had an impact or affected the internal validity. 3. If O2 and O5 are equal, then the internal validity is not affected by testing effects. 4. Compare the scores of O6 with O1 and O3 and if all the scores are similar, maturation and history effects have not been a problem. The Solomon four group experimental design guarantee the maximum internal validity. However, the cost of conducting this kind of experiment is high. Major threats to internal validity in different experimental designs when members are selected and assigned are: Types of experimental designs 1. Pre test and post test with one experimental group only 2. Post test only with one experimental and one control group Major threats to internal validity Testing, history and maturation Maturation 41 3. Pre test and post test with one experimental and one control group Mortality 4. Solomon four group design Mortality Double blind study When the subjects in the experimental and control groups to whom drugs administered are kept unaware, it is called as blind studies. When both experimental designs are blinded, such studies are called as ‘double blinded studies’. As there was no tampering with the treatment, such experimental designs are the least biases. Example When an outside agency handles the entire process of experiment, they know who is true vs. the ‘placebo’ treatment. (Placebo means, medicine prescribed for the mental benefit of the patient rather than for any physical effect) Ex-post Facto Design The study followed much later to the treatment effect is called ex-post facto design. Example A training programme was introduced in an organization about two years ago. Some employees attended and some did not. Performance data is collected now only. Simulation It is an alternative to lab and field experimentation. It is a model building technique to determine the effect of changes. Computer based simulations are becoming popular in business research. Simulation closely represents the natural environment. Simulation lies between lab and field experiment. It is artificially created but not far from reality. Participants are exposed to real world experience. Example A real office is set-up. Members are assigned the role of director manager etc. randomly. The researcher would retain the control and members left free to operate as in a real office. Data on the independent variable can be obtained through observation, videotaping, audio recording, interviews or questionnaires. Casual relationship can be tested by manipulation and control in a possible simulation. We can conduct two types of simulation. They are: 42 1. The nature and timing of simulated events are totally determined by the researcher (experimental simulation). 2. The course of activities is at least partly governed by the reaction of the participants to the various stimuli as they interact among themselves. Examples Flight simulators, driving simulators or nuclear simulators are some of the examples. Unethical in experimental design research The following are the unethical practices in experimental research design that needed to be avoided. Putting pressure on individuals to participate in experiments Giving menial tasks (unworthy) and asking (demeaning) (beneath the dignity) questions, which will diminish the members self respect Deceiving subjects (members) by deliberately (wantonly) misleading them as to the true purpose of research Exposing participants to physical or mental stress Not allowing the subjects to withdraw from the research when they want Using the results to the disadvantages of participants or for the purposes not to their liking Not explaining the procedure to be followed Exposing the respondents to hazardous and unsafe environments as we saw earlier in the case of Johns Hopkins University Not debriefing participants fully and accurately after the experiment is over Not preserving the privacy and confidentiality of the information given by the participants Withholding benefits from control groups --- 43 Sampling Census and Sample Survey / Determining Sample Design All the items under consideration in any field of enquiry constitute a ‘universe’ or ‘population’. When all the items in the population are taken into account it is known as census inquiry. There is no element of chance left and highest accuracy is obtained when all the items are covered in an enquiry. In practice it may not be true. Slightest bias in such an enquiry will become larger and larger as the number of observation increases. There is no method to check the element of bias or its extent unless through a resurvey or use sample checks. Again, the type of inquiry involves a great deal of time, money and energy. Census inquiry is not possible under many circumstances. Only the Government will be able to complete the exercise. Govt. conducts census survey once in ten years. It is not possible to examine every item in the population. Sometimes, it is possible to obtain sufficiently accurate results by studying only a part of total population. In such cases there is no utility of census survey. When the universe is small there is no use resorting to sample survey. Consideration of time and cost, almost invariably, lead to selection of respondents, i.e. selection of only few items. The respondents selected should be representative of total population (N) as possible to produce a miniature cross-section. The selected respondents constitute ‘sample’ (n units) and the selection process is called ‘selection technique’. The survey so conducted is known as ‘sample survey’. The researcher must plan how a sample should be selected and what size such sample should be. Example: Blood testing is done only on sample basis. We select only few items from the universe to study. Sample design is a definite plan determined before any data are actually collected. A plan to select 10 out of a 150 drugstores in a certain way constitutes a sample design. Samples can be either probability samples or non-probability samples. With probability sample each element has a known probability of being included in the sample. But the non-probability samples do not allow the researcher to determine this probability. Implications of Sample Design A sample design is a definite plan for obtaining a sample from a given population. It is the technique or the procedure the researcher would adopt in selecting items for the sample. Sample design may as well lay down the number of items to be included in the sample i.e. the size of the sample. Sample design is determined before data are collected. There are many sample designs from which the researcher can choose. Some designs are relatively more precise and easier to apply than others. Researcher must select/prepare a sample design which should be reliable and appropriate for his research study. 44 Steps in Sample Design While developing a sample design, the researcher must pay attention to the following points: 1. Type of universe Define the objects, technically called universe. Universe can be finite or infinite (the population of a city, number of workers in a factory etc.). In finite, the number of items is certain and infinite there is no idea about total number of items (number of stars in the sky, listeners of a specific radio programme, throwing of a dice etc.). 2. Sampling unit Take decision on sampling unit before selecting the sample. It can be geographical (state, district, village etc.) or a construction unit (house, flat etc.,) or it may be a social unit (family, club, school etc.) or it may be an individual. The researcher has to decide and select one or more of such units for his study. 3. Source list Sample has to be drawn from source list known as ‘sampling frame’. It contains the names of all items in the universe in the case of finite universe. Researcher has to prepare the source list, if it is not available. Source list must be comprehensive, correct reliable and appropriate. Source list should be representative of the population as far as possible. 4. Size of sample This refers to the number of items selected from the universe. It should not be large or small but should be optimum. Optimum sample should conform to efficiency, representative ness, reliability and flexibility. Researcher must determine the desired precision as an acceptable confidence level of the estimate. The size of population variance needs to be considered as in case of larger variance usually a bigger sample is required. Budgetary constraint has to be taken into consideration while deciding the sample size. 5. Parameters of interest One must consider the specific population parameters which are of interest. We may be interested in estimating the proportions of person with some characteristic in the population or we may be interested in some average or other measure concerning the population. There may be some important sub-groups in the population; we would like to make estimates. All these have strong impact on the sample design we would accept. 6. Budgetary constraint Cost considerations impact upon not only the size of sample but also to the type of sample. This fact even leads to the use of non-probability sample. 7. Sampling procedure There are several sample designs out of which the researcher must choose one for his study. He must select a design for a given sample size and for a given cost with a smaller sampling error. 45 Characteristics of a Good Sampling Design 1. 2. 3. 4. 5. 6. The characteristics of a good sample design are listed below: Sample design must result in a truly representative sample. Sample design must be such with results in a small sampling error. Sample design must be viable in the context of funds available for the research study. Sample design must be such so that systematic bias can be controlled in a better way. Sample should be such that the results of the sample study can be applied, in general, for the universe with a reasonable level of confidence. Different Types of Sample Design Probability samples include simple random sampling, systematic sampling, stratified sampling, and cluster/area sampling. Non-probability samples include convenience sampling, purposive sampling (judgment sampling and quota sampling) techniques. Sampling methods Non-probability Sampling Purposive Sampling Judgment Sampling Probability Sampling Convenience Sampling Quota Sampling Random Sampling Stratified Random Sampling Proportionate stratified Random Sampling Systematic Random Sampling Cluster Sampling Disproportionate Stratified Random Sampling Unrestricted or Simple Random sampling It is known as chance sampling or probability sampling. Each and every item in the population has an equal chance of inclusion in the sample. Each one of the possible samples, in case of finite universe, has the same probability of being selected. Example: Selection of 300 samples from a universe of 15000 items. Conduct lottery. Using random number table is another method. Each item is assigned a five digit number from 1 to 15000 and 300 five digit numbers are selected using the random number table. Select some random starting point and use a systematic pattern 46 to proceed through the table. As the numbers placed in the table are in a completely random fashioned, the resulting sample is random. Here, each item has an equal probability of being selected. In case of infinite population, the selection of each item in random sample is controlled by the same probability and the successive selection are independent of one another. Every element in the population has a known or equal chance of selection is known as unrestricted probability sampling design or simple random sampling. For example, there are 1000 pieces in the population and we need 100 samples. You drop all the 1000 elements into a hat and draw 100 of those elements from the hat by closing your eyes. The first element will have a 1/1000 chance of drawing and the second element will have a chance of 2/999 and so on. That means, the probability of any one of the element being chosen is 1 in the number of population and each element has the same or equal probability of being chosen. Similarly, the distribution patterns of the characteristics are also distributed in the subjects drawn from the sample. This simple random sampling has the least bias and offers more generalisability. However, the sampling process is cumbersome and expensive. Entirely updated population list may not be available. Therefore, other probability designs are considered. Restricted or Complex Probability Sampling The restricted or complex probability sampling procedure provides a viable and more efficient alternative to unrestricted or random sampling design as discussed above. We can get more information by using some of the complex probability sampling procedures than unrestricted or random sampling design. The five most complex probability sampling are discussed below: 1. 2. 3. 4. 5. systematic sampling stratified random sampling cluster sampling area sampling and double sampling Systematic Random Sampling Sometimes, it is more practical way of sampling to select every 15th name on a list or 10th house on the one side of a street etc. This type of sampling is known as systematic sampling. An element of randomness is used to pick up the given unit with which to start. The selection process starts by picking some starting point in the list and every nth element is selected until the desired number is secured. It is drawing the nth element between 1 and n. Please look into the following example: You want a sample of 25 houses from a total population of 250 houses. You could sample every 10th house from a random number 1 to 10. If the random number is 10, the houses numbered 10, 20, 30, 40, and so on until 25 houses are selected. The problem in this type of sampling is systematic bias. For instance, every 10th house is a corner house. You want to study to control the noise pollution by using the appropriate filtering materials. The corner houses are not 47 exposed to as much noise of the houses in between. It is a systematic bias built in and the researcher might be drawing wrong conclusions from such study. Researcher has to ensure whether the systematic sampling is appropriate before deciding on it. We use systematic sampling design for market surveys, consumer attitude survey and the like. The telephone directory is the population frame for this sampling design. Stratified Random Sampling We apply stratified random sampling technique to obtain a representative sample, if the population from which a sample is to be drawn, does not constitute a homogeneous group. In this technique, the population is stratified into a number of non-overlapping subpopulation or strata. Sample items are selected from each stratum based on simple random sampling, i.e. first stratification and then simple random sampling is known as stratified random sampling. Sampling helps to estimate population parameters. There may subgroups of elements within the population, which may have different parameters on the variable of interest to the researcher. For example, the entire organization will form the population for study of training needs. The extent, quality, and intensity of training requirement may vary to the middle level manages, lower level managers, first line supervisors, clerical workers etc., The director of a company can develop useful and meaningful training program as per the requirement of each sub group mentioned above. We collect the data to assess the needs at each sub group level in the population. The stratified random sampling used at each sub group level will be the aggregate for the unit analysis at the group level. We divide the population into mutually exclusive groups that are relevant, appropriate, and meaningful, relevant to the context of the study. Example: A CEO of a company may be concerned about the motivational levels and high absenteeism rates amongst the employees in the company. It is sensible to stratify the population of the employees according to their job levels and collect the data. We may find to our surprise, that middle level managers were not motivated. This kind of information will help the CEO to focus and devise better methods to motivate this group. It is not possible to trace the differences within the group without stratified random sampling procedure. Stratification helps to answer the following research questions: 1. Are the machinists more accident prone than clerical workers? 2. Are Hispanics more loyal to the organization than Native Americans are? The common examples of the use of stratification as a sampling technique are: Life stages (child, adolescent, adult etc.,), income levels, buying patterns are some examples for stratification (arranging) of customers. We can stratify the companies according to size, industry, profits etc. 48 Stratification is an efficient research sampling design. It provides more information with the sample size. Study of consumer preferences for a product, the stratification on the population could be by geographical areas, market segments, consumer’s age, consumer’s gender, or various combinations of these. Stratification ensures homogeneity within each stratum (with very few differences or dispersions on the variable of interest within each stratum), but heterogeneity (variability) between strata. It means there will be more between group’s differences than withingroup differences. Cluster Sampling It involves grouping the population and selecting the groups or clusters rather than the individual elements for inclusion in the sample. Example: A departmental store has issued 15000 credit cards and would like to sample its credit cardholders. Say, the sample size has to be kept at 450. 100 clusters of 150 cardholders each could be formed. Three clusters can be selected randomly. The sample size should be larger than the simple random sample to ensure the same level of accuracy, because in cluster sampling there is procedural potential for order bias and other sources of errors usually accentuate. It involves the division of the population into convenient clusters, randomly choosing the required number of clusters as sample subjects, and investigating all the elements in each of the randomly chosen clusters. We do cluster sampling in several stages and we call it as multistage cluster sampling. Example: We are doing a national survey on bank deposits. We will select urban, semi-urban, and rural geographical locations for study of cluster sampling. Next stage, we will choose particular areas in each of these locations. At third stage, we choose banks within each area, i.e. until we have reached the final stage of breakdown for sample units, when we will sample every member in those units. Area Sampling It is quite close to cluster sampling. It is used when the total geographical area is very big. Total area is divided into a number of smaller non-overlapping areas, generally called geographical clusters. Then a number of these smaller areas are randomly selected. All the units in the selected small area are included in the sample. Area sampling is helpful when we do not have the list of the population concerned. It makes the field interviewing more efficient because the interviewer can conduct many interviews at each location. The area sampling is a form of cluster sampling within an identifiable geographic area. Example: You want to open a 24-hour convenience store in a particular part of the town. You have to do a sampling survey, to know about the needs of the consumers in that area to meet their requirement or purchase. Area sampling is less expensive. It is not dependent on a population frame. A city map showing the blocks of the city is sufficient. Double Sampling We use a sample in a study to collect some preliminary information of interest. We use a subsample of this primary sample to examine the matter in more detail. We call it as double sampling. Example: A structured interview may indicate that the respondents of a sub-group 49 have more in depth knowledge of the problem in an organization. When we interview these respondents and ask additional questions, we adopt double sampling procedure. Proportionate and Disproportionate Stratified Random Sampling We can draw a sample of members from each stratum using either simple random sampling or systematic sampling procedure. The subjects drawn from each stratum can be proportionate or disproportionate in the number of elements in the stratum. We explain the proportionate and the disproportionate stratified sampling below: Job level Top management Middle-level mgt Lower-level mgt Supervisors Clerks Secretaries Total Number of Elements 10 30 50 100 500 20 710 Number of subjects in the sample Proportionate Sampling Disproportionate (20% of the elements) Sampling 2 6 10 20 100 4 142 7 15 20 30 60 10 142 Proportionate Stratified Random Sampling: Take 20% of members from each stratum. That means the members represented in the sample from each stratum will be proportionate to the total number of elements in the respective strata. Please refer to the above table. Disproportionate Stratified Random Sampling: The researcher may be concerned that few members from the top management and more members from the lower levels will not truly reflect how all members in a proportionate stratified random sampling will respond. He may use a disproportionate stratified random sampling procedure, by altering the number of subjects from each stratum and keeping the sample size. Please refer to the above table. When some stratum or strata are too small or too large, we make disproportionate sampling decisions. Example: Educational levels among supervisors may range from elementary school to masters degrees. We take more samples at supervisory level. We do disproportionate sampling, when it is easier, simpler and less expensive to collect data from one or more strata than from others. The proportionate and the disproportionate stratified samples is more efficient than the simple random sampling design because, each important segment of the population is better represented and more valuable and differentiated information is obtained with respect to each group, for the sample size. Single stage or Multi-stage Sampling his is further development of the idea of cluster sampling. This technique is used for big enquiries or covering the geographical area of the entire country. Under multi-stage sampling the 50 first stage may be to select large primary sampling units such as states, then districts, then towns, and finally certain families within towns. If the technique of random sampling is applied at all stages, the sampling procedure is described as multi-stage random sampling. Sampling helps to estimate population parameters. There may subgroups of elements within the population, which may have different parameters on the variable of interest to the researcher. For example, the entire organization will form the population for study of training needs. The extent, quality, and intensity of training requirement may vary to the middle level manages, lower level managers, first line supervisors, clerical workers etc., The director of a company can develop useful and meaningful training program as per the requirement of each sub group mentioned above. We collect the data to assess the needs at each sub group level in the population. The stratified random sampling used at each sub group level will be the aggregate for the unit analysis at the group level. Review of probability Sampling Designs There are two basic probability-sampling plans. 1. the unrestricted or simple random sampling and 2. the restricted or complex probability sampling plans Every element in the population has a known and equal chance of selection as a subject, in the simple random sampling. The complex probability plan consists of five different sampling designs. The cluster sampling is convenient, least expensive and dependable. As there is no heterogeneity among the elements, does not offer much efficiency in terms of precision or confidence in the results. The stratified random sampling is precise and efficient. It offers precise and detailed information. The systematic sampling has the built-in hazard for possible systematic bias. Area sampling is a popular form of cluster sampling. We use double sampling when we want additional information by using the primary sample to examine the subject of interest in more detail. Non probability Sampling In non-probability sampling, the elements in the population do not attach any probability for selection as sample subjects, because we can not generalize the findings of the surveys to the population. Sometimes the researchers are less concerned about generalize to the population in view of obtaining certain preliminary information quickly and inexpensively and may resort to non-probability sampling. Some times the non-probability sampling could be the only way to obtain the data. We shall discuss about this later. The non probability sampling plans are more dependable, than others and can lead to potentially useful information in respect of the population. The non-probability sampling designs fall into the following categories: 1. convenience sampling and 2. purposive sampling 51 Convenience Sampling Convenience sampling is the collection of information from members of the population conveniently available. We call it as convenience sampling. We can conduct convenience sampling at a shopping mall, visited by many shoppers to find out, whether people prefer one product to another. ‘Pepsi challenge’ contest was administered on a ‘convenience’ sampling basis. We use this kind of sampling during the exploratory phase of a research project and it is the best way to get information quickly and efficiently. Another example, Five officers visited the competitors’ demonstration of a new product, informed the Vice President of their company about the move of the competitor. It helped the Vice President to position his product suitably in the market. Here the Vice president gathered the information conveniently from the officers visited the competitor’s show room. Purposive Sampling Sometimes we may have to obtain information from specific target group. The sampling will be restricted to specific type of peoples who can provide the desired information. Here the restricted people alone have the information or can provide as per the criteria provided by the researcher. We call this type of sampling design as purposive sampling, which consists of the following types: 1. judgment sampling and 2. quota sampling Judgment Sampling Judgment sampling design is used when only a limited number or category of people who can provide the same. Example: You want to find out what made the women managers to make to the top, only the people moved to the top position will have the information to answer. However, the judgment sampling will reduce the chances of generalizing as we are using the conveniently available experts and there is no other option too. As very knowledgeable people are included in the sample, the data source will be rich. The researcher has to take special efforts to locate and access to the kind of people referred above. Quota Sampling We fix the quota for each sub group based on the total numbers of each group in the population. As it is a non probability sampling plan, the results are not generalizable to the population. We take predetermined proportion of people as samples from different groups, but on a convenience basis. Example: We may surmise (without having any evidence), that work attitude of blue-collar workers is different from white-collar workers in an organization. Suppose, we have to interview in all 30 people and the ratio of blue-collar workers and white-collar workers is 60:40, the quota for sampling shall be the first 18 blue-collar workers and 12 white-collar workers respectively.. 52 The sample may not be the representative of the population and the generalization will be restricted (limited). Quota sampling is necessary when there is no adequate representation of sample in the sub set of the population. Quota samples are stratified samples from which we select the subjects randomly. We use the quota sampling to know the buying behavior of various groups, to get a feel of how employees from different nationals perceive the organizational culture, etc. If quota sampling offers certain useful information, based on that further investigation, (probability sampling), can be done. A probability sampling might indicate new areas of research, and non-probability sampling designs can be used to explore their feasibility. Collecting the Data The data that is available in hand is inadequate. It is necessary to collect appropriate data. Collection of data depends upon factors like money costs, time and other resources at the disposal of the researcher. Primary data can be collected through experiment or survey. In an experiment the researcher observes some quantitative measurement or data and examines them with the truth contained in his hypothesis. In this context Dr.A.L.Bowley very aptly remarks that in collection of statistical data common sense is the chief requisite and experience is the chief teacher. In respect of survey, data can be collected by any one or more of the following ways: By observation Investigator collects information by observation, without interviewing. The information collected relates to what is currently happening and not past behavior or future intentions or attitudes of the respondents. The method is very expensive. Only limited information can be collected. This method is not suitable where large samples are involved. Through personal interview The investigator follows a rigid procedure and seeks answers to a set of preconceived questions through personal interviews. The questions are asked in a structured way and the output depends upon the ability of the interviewer to a larger extent. Through telephone interviews Information is collected through telephone. It is not a very widely used method. It plays an important role in industrial surveys when the survey has to be completed in a very short time. By mailing of questionnaires The researcher and the respondents come into contact. Questionnaire is mailed to the respondents with the request to return duly completed. It is extensively used method in economic and 53 business surveys. A pilot study for testing the questionnaire is conducted before applying this method to overcome weaknesses, if any, of the questionnaire. Questionnaire must be prepared carefully to collect the information effectively. Through schedules Enumerators are appointed and given training. They are provided with schedules with relevant questions. They go to the respondents, collect and fill-up the data on the basis of replies given. This method depends mostly on the capability of enumerators. Occasional check on the enumerators may ensure sincere work. Execution of the Project It is a very important step in research process. The data to be collected would be adequate and dependable if the project proceeds on correct lines. Project must be executed systematically and in time. Data can be machine processed in case of structured questionnaires. Questions as well the possible answers may be coded. In case the data to be collected through interviews, proper selection and training of the interviewers. Training must be supported with instruction manuals explaining the job of interviewers at each step. Occasional field check on the interviewers is necessary. Keep a careful watch on unanticipated factors so that the survey is as realistic as possible. In other words, ensure that the survey is under statistical control so that the collected information conforms to the predefined standard of accuracy. Take suitable measures to tackle when respondents do not cooperate. Analysis of Data Analysis of data includes, establishment of categories, application of these categories to raw data through coding, tabulation and drawing statistical inferences. Condense the unwieldy data into few manageable groups and tables for further analysis. Coding operation is done at this stage to transform categories of data into symbols that may be tabulated and counted. Editing is the procedure to improve the data for coding. Now the stage is ready for tabulation. Tabulation is a technical procedure where in the classified data are put in the form of tables. A great deal of data in respect of large enquiries can be tabulated by computers to save time as well study large number of variables affecting a problem simultaneously. We can use statistical tests to establish whether two different mean values are significantly different (a real one) or the difference is just a matter of chance (random fluctuations). If the difference happens to be real, the inference will be that the two samples come from different universes and if the difference is due to chance, the conclusion would be that the two samples belong to the same universe. Hypothesis Testing Now the researcher is in a position to test the hypothesis, if any, he had formulated earlier to find out whether the facts collected supports the hypothesis or happen to be contrary. The researcher can test the hypothesis using one or more tests, viz. Chi square test, t-test, F-test, developed by 54 statisticians for testing purposes. Hypothesis testing will result in either accepting or rejecting it. If no hypothesis to start with, generalization established on the basis of data may be stated as hypothesis to be tested by subsequent researches in times to come. Generalization and Interpretation Researcher can generalize i.e. build a theory, when hypothesis is tested and upheld several times. If the researcher has no hypothesis to start with, he might seek explain his findings on the basis of some theory. It is known as interpretation. The process of interpretations may quite often trigger off new questions which in turn may lead to further researches. Preparation of the Report or the Thesis The report should be carefully written duly taking into the following: 1. The layout of the report should be as follows: a. The preliminary pages b. The main text and c. The end matter In its preliminary pages the report should carry title and date followed by acknowledgements and foreword. Then there should be table of contents followed by a list of graphs and charts, if any, given in the report. The main text of the report should contain the following: a. Introduction It should contain a clear statement of the objective of the research and an explanation of the methodology adopted in accomplishing the research. The scope of the study along with various limitations should be well stated in this part. b. Summary of findings Incorporate a statement of findings and recommendations in non-technical language. Summarize findings if they are extensive. c. Main report The main body of the report should be presented in logical sequence and broken down into readily identifiable sections. 55 d. Conclusion 1. The researcher has to, at the end of the report; sum up the results of his research clearly and precisely. 2. Report should be written in a concise and objective style in simple language avoiding vague expressions such as ‘it seems’, ‘there may be’, and the like. 3. Charts and illustrations in the main report should be used only if they present the information more clearly and forcibly. 4. Calculated ‘confidence limits’ must be mentioned and the various constraints experienced in conducting research operations may as well be stated. Criteria of Good Research Scientific research should satisfy the following: 1. The purpose of the research should be clearly defined and common concepts used. 2. The research procedure used should be described in sufficient detail to permit another researcher to repeat the research for further advancement, keeping the continuity of what has already been attained. 3. The procedural design of the research should be carefully planned to yield results that are as objective as possible. 4. The researcher should report with complete frankness. flaws in procedural design and estimate their effects upon the findings. 5. The analysis of data should be sufficiently adequate to reveal its significance and the methods of analysis used should be appropriate. The validity and reliability of the data should be checked carefully. 6. Conclusions should be confined to those justified by the data of the research and limited to those for which the data provide an adequate basis. 7. Greater confidence in research is warranted if the researcher is experienced, has a good reputation in research and is a person of integrity. The good qualities of research are: 1. Good research is systematic: Systematic characteristic of the research does not rule out creative thinking but certainly rejects the use of guessing and intuition in arriving at conclusions. 2. Good research is logical: The research is guided by the rules of logical reasoning and logical process of induction and deduction are of great value in carrying out research. Induction is the process of reasoning from a part to the whole. The deduction is the process of reasoning from some premise to a conclusion which follows from that very premise. In fact, logical reasoning makes research more meaningful in the context of decision making. 3. Good research is empirical: The research is related basically to one or more aspects of a real situation and deals with concrete data that provides a basis for external validity to research results. 4. Good research is replicable: This characteristic allows research results to be verified by replicating the study and thereby building a sound basis for decisions. 56 Problems encountered by researchers in India: 1. There is lack of scientific training in the methodology of research. 2. Insufficient interaction between the university research departments and business establishments, govt. depts., and research institutions on the other hand. 3. Need to generate confidence that data collected from business unit will not be misused. 4. Research studies overlap one another often for want of adequate information. 5. No code of conduct exists for researchers. 6. Researchers face difficulty of adequate and timely secretarial assistance. 7. Library management and functioning is not satisfactory at many places. 8. Many libraries are unable to get copies of old and new Acts/Rules, reports and other government publications in time. 9. Difficulty of timely availability of published data. 10. Problem relating to process of data collection and related things. 57 DATA COLLECTION METHOD INTRODUCTION We will study about the various sources of data and what data we can collect for the purpose of analysis, testing the hypothesis, and answering the research questions. The sources of data and the way in which it is collected will make a great difference to the rigor and effectiveness of the project reports. SOURCES OF DATA Primary data: First-hand information obtained by the researcher on the variables that are required for the specific purpose of the study. Example: individuals, focus groups, panels of respondents specifically set-up by the researcher to obtain opinions on specific issues or some unobtrusive sources (unattractive waste materials) such as trash can (dust bin). We can consider internet as a primary data source. Secondary data: Information gathered from the sources already existing, viz. company records, or archives, government publications, industry analysis, web sites, internet etc. Primary sources of data: 1. Individuals 2. Focus groups 3. Panels 4. Unobtrusive methods Individuals: Individuals provide information when interviewed, administered questionnaires or observed. Focus Groups: It consists of 8 to 10 members. A moderator will lead the discussions for about 2 hours on a particular topic, concept or product. Expertise is the main criteria to choose members. We aim to focus sessions to obtain the respondents' impressions, opinions, or interpretations on the topic. The respondents will be spontaneous in their responses and reflect the genuine ideas, opinions and feelings, as the session is unstructured. Focus groups are relatively inexpensive. The data given by them in a shorter period is reliable. Examples are Computer specialists (to discuss matters relating to computing), women with children (to help working mothers). The role of the moderator: The moderator introduces the topic, observes and takes notes. He will not be the part of the group. He steers the discussions, keep the members on track and persuade the group to obtain the relevant information. He ensures that all members are participating. He does not allow any member to dominate. He will notice the verbal statements as well the nonverbal cues (signals) of the members. 58 The nature of data obtained through focus groups: The focus group is homogeneous (alike). It is least expensive of all data collection method. We can analyze the data collected quickly. The data obtained provide only qualitative information and not quantitative. We cannot consider their opinion truly representative, as the selected members do not reflect the opinions of the populations at large. The focus group plays an important role while collecting the exploratory information as a basis for further scientific research. When animated discussions takes place, the serendipitous (the occurrence of events by chance in a fortunate way) flow of new ideas help the researchers to obtain valuable information. We use the focus groups for the following purposes: 1. exploratory studies 2. making generalizations based on the information generated by them and 3. conducting sample surveys Video Conferencing: It is gathering information from different groups in distant places. The moderator relays instant messages. We can record the nonverbal cues (signals) and gestures accurately. E-mail, web sites and Internet chat rooms facilitate focus group sessions as well. Panels: Panels meet more than once. Focus groups meet for one-time-session. We use the panels to study changes over a period (Example: proposed advertisement of a given brand). The continuing set of experts serves as sample base. When the research uses the experts, we call it as ‘panel study’. E.g., The design of Nielson television index provides the size and nature of the audience for individual television program by connecting audiometer instruments. Static and dynamic panels: 1. static ( same members in the panel over a period of time) or 2. dynamic (members change from time to time when the study is in progress). Panel is the source of direct information. The advantages static panel – offers a good sensitive measurement of changes that are taking place between two points of time. Disadvantage – Panel members become sensitive and no longer represent the opinion of the population. Members can also drop out from time to time, raising issues of bias due to mortality. Unobtrusive or Trace Measures: These measures do not involve people. Their uses are important to research. e.g. are: wear and tear of journals in a university indicates their popularity. cheques exposed to ultra violet rays indicate to the extent of frauds or forgery company records disclose lot of personal information about employees 59 number of different brands of soft drink found in the trash can also indicate the measure of their consumption levels. Secondary sources of data: It is information collected other than the researcher conducting the current study. Such data can be internal or external to the organization. The researcher can access data through the internet or perusal of recorded or published information. He can use the secondary data to forecast sales by constructing models based on the past sales figures and through extrapolation. The secondary data saves time and cost in acquiring the information. The drawback is it may become obsolete and may not meet the specific needs of the given situation. Therefore, it is important to refer to sources that offer current and up-to-date information. DATA COLLECTION METHOD: The three main data collection methods are: 1. interviewing 2. administering questionnaires and 3. observing people and phenomena (a fact or situation that is observed to exist or happen) Interviewing has the advantage of flexibility in terms of adapting, adopting and changing questions as the researcher proceeds with the interview. Questionnaires have the advantage of obtaining data more efficiently in terms of researcher’s time, energy, and costs. Unobtrusive methods of data collection from company records have the advantage of accuracy. Modern technology plays an important role in shaping the data collection methods. e.g. The researchers use computer-assisted survey, computer assisted telephone interviewing, interactive electronic surveys, administering questionnaires through electronic mail (e-mail) etc., to gathering data. The choice of data collection methods depends on the facilities available, the degree of accuracy required, the expertise of the researcher, the time span of the study, and other costs and resources associated with and available for data gathering. PART I: INTERVIEWING We interview respondents to obtain information on the issues of interest. The interviews can be unstructured or structured and conducted either face to face or by telephone or online. Unstructured interviews: The interviewer will not have a planned sequence of questions. The objective is to unearth some preliminary issues. The researcher decides what variable needs further investigation. The 60 manager may have a vague (not certain or definite) idea of certain changes taking place. Such situations call for unstructured interviews. To begin with, the interviewer asks only broad, openended questions. Based on the response the researcher will form the perceptions of the individuals. The interviewer asks direct questions to top and middle level managers whereas the lower level employees broad, open-ended questions as follows. “Tell me something about your unit or department, and perhaps even the organization as a whole, in terms of work, employees, and whatever else you think is important.” Some people will give elaborate reply to the above questions, and others may say that everything is fine. As managers and researchers, we should listen to the very important messages, conveyed very casually by the respondent and identify the critical topics. When the respondents are answering in one or two words, we can frame questions as follows: “I would like to know something about your job. Please describe to me in detail the things you do on your job on a typical day, from eight in the morning to four in the afternoon.” Several questions can follow-up to the answer. Compared to other units in this organization, what are the strengths and weaknesses of your unit?” If you would like to have a problem solved in your unit, or a bottleneck eliminated, or something attended to that blocks your effectiveness, what would that be?” The respondents may say ‘fine’ and have no problems. The researcher then can ask: “That is great! Tell me what contributes to this effectiveness of your unit, because most other organizations usually experience several difficulties.” The interviewer asks the lower level employees very broad questions relating to their jobs, work environment, satisfactions and dissatisfactions at the work place, and the like as follows: What do you like about working here? If you were to tell me what aspects of your job, you like and what you do not, or what would they be? Tell me something about the reward system in this place. If you were offered a similar job elsewhere, how willing would you be to take it and why? If I were to seek employment here and request you describe your unit to me as a new comer, what would you say? 61 The researcher will know the variables that require greater focus after conducting the unstructured interviews several times at all levels and call for more in-depth information. The researcher will conduct structured interviews for the variables identified. Structured Interviews: The researcher conducts structured interviews on the information needed. Interviewer will have a list of predetermined (pre-arranged) questions. Questions may be asked to the respondents either personally, or through telephone or through a medium of a PC. The researcher designs the questions based on the variables surfaced in the unstructured interviews and considered relevant to the problem. Questions will be uniform to all the respondents. The researcher may take a lead from the respondent in the course of interview and ask other relevant questions conforming to the goal of the interview. New factors identified, resulting the deeper understanding of the problem. Visual aids: We use pictures, line drawings, cards, etc., in interviews. The interviewees indicate their responses to questions by seeing the visuals. Visuals overcome the difficulty of asking questions that would cause and embarrass the feelings of the interviewees. e.g. are: It helps to identify the likes and the dislikes of the customers to different types of packaging, forms of advertisement etc. Visual of painting and drawing is useful when focusing children in the marketing research. The researcher would stop the interview after conducting sufficient number of interviews and obtaining adequate information. He will then tabulate the data and analyze. He will then describe the phenomena (observation of a fact or situation that exists or happens), or quantify them, or identify the specific problem and evolve a theory of the factors that influence the problem or find answers to the research problem. Training Interviewers: We require a team of interviewers They need to be thoroughly briefed about the research and trained in how to start the interview, how to proceed with questions, how to motivate respondents to answer, what to look for in the answers and how to close the interview. Some tips to follow in interviewing: The information obtained in the interview must be free from bias. He must listen attentively to the interviewee, exercise tact in questioning, repeating and/or clarifying the questions posed and express the meanings of different word used. It is important to record the responses accurately. He must not influence the responses of the interviewees, and must not distort or falsifying the information received. Establishing credibility and rapport, and motivating individuals to respond: The researcher must possess the qualities of knowledge, skills, ability, confidence, articulateness and enthusiasm in order to establish credibility with the hiring organization. He must keep the interviewee at ease to obtain truthful answers without fear of adverse consequences. Establishing 62 rapport with the respondents especially, the lower level employees is not easy. The researcher must tactfully make it clear to the respondents that his purpose is to understand the true status of affairs in the organization and do not intend to take sides. He should not identify the individuals in the process of surveys. This will help encourage the respondents to feel secure in responding. The researcher can gain confidence and rapport by being pleasant, sincere, sensitive and nonevaluative. He should evince genuine interest in responses and reduce any anxieties, fears, suspicion or tension sensed in the situation to help the respondents feel more comfortable with the researchers. He must motivate the respondents that their contributions in the process will improve the quality of life significantly. THE QUESTIONING TECHNIQUE Funneling: It is preferable to ask open-ended questions while beginning an unstructured interview. For example: What are some of your feelings about working for this organization? In response to the broad question, further questions can be progressively focused, relevant to the situation. The process of changing from the broad to narrow themes is called funneling technique. Unbiased Questions: There must not be any bias in asking questions. For example: It would be better to ask “Tell me how you experience your job” than “Boy, the work you do must be really boring; let me hear how you experience it”. Clarifying issues: The researcher must make sure that he understands the issues, and the intention of the respondents from their replies. For example: The respondent may reply, “Facilities here are really poor; we often have to continue working even when we are dying of thirsty”. Then the researcher might ask if there is no drinking water available in the building. The respondent’s reply could be “though there is water available in the building, would like to have one on the side of his work area”. Helping the respondents to think through issues: The respondent may say ‘I don’t know’. The researcher should ask paired questions “whether you prefer to serve the customer or do some filing work”. If the reply is to serve the customer, 63 the researcher can ask further paired-choice question again. In the process, the respondent can sort out which aspects of the work he likes. Taking notes: The researcher must take notes during the interview or as soon as it is over. He cannot rely on memory. Taped or video recording might bias the respondent because they are afraid of disclosing their identity. The researcher must obtain the respondent’s permission for recordings. Review of Interviewing: We conduct unstructured interviews to obtain definite ideas. Structured interviews give more in depth information about specific variables of interest. The interviewer must establish rapport with the respondent and ask unbiased questions. The face-to-face interview and the telephone interview have their own advantages and disadvantages and both have their uses in different circumstances. Computer assisted interviews shall be important mode of data collection in the future. PART II: QUESTIONNAIRES Questionnaires are efficient data collection mechanism. The researcher knows what is exactly required and how to measure the variables. The researcher administers the questionnaires to the respondents personally, mails to the respondents, or electronically distributes. Personally Administered Questionnaires: The group of employees at the work place will respond to the questionnaire. The advantages are: You can collect responses to the questionnaire in a shortest span of time. You can clarify the doubts on any question on the spot. Motivate the respondents to offer their frank answers. Less expensive and consume less time than interviewing taking into account of administering the questionnaire to large number of individuals. Some companies do not allow employees to complete the questionnaire during the working hours. In such instances, you can give blank questionnaires and collect within a short time limit. You can provide scanner sheets for the multiple choices for direct entry into the computer. Mail Questionnaires: We can cover a wide geographical area. A 30% response rate is acceptable. The disadvantage: we may not clarify the doubts of the respondents. If the responding to the survey is low, the survey may not represent the population. We can improve the response by providing some incentive and self addressed, stamped return envelopes and a brief questionnaire. You can improve the response better, if you notify the respondents about the survey in advance and involve a reputed firm in the survey with their covering letter. 64 Field surveys, comparative studies and experimental designs often use questionnaires to measure the variables of interest. We must know how to design them better. Guidelines for Questionnaire Design: Please refer to figure 10.1. Sound questionnaire design must focus on the following important issues to reduce the bias in the research. 1. the wording of questions 2. planning of issues of how the variables will be categorized, scaled and coded after receipt of responses 3. general appearance of the questionnaire PRINCIPLES OF WORDING: The principles of wording refer to the following factors: 1. 2. 3. 4. 5. the appropriateness of the contents of the questions how questions are worded and the level of sophistication of the language used the type and form of questions asked the sequencing of the questions the personal data sought from the respondents Content and purpose of the questions: The questions should tap the dimensions and elements of the concept when the variables are of subjective in nature (e.g. satisfaction, involvement), where respondents’ beliefs, perceptions, and attitudes are to be measured. A single direct question (one that has an ordinal-scaled set of categories) is sufficient to tap the objective variables such as age and educational levels of the respondents. Consider the purpose of each question carefully to measure the variables adequately and do not ask any superfluous question Languages and wording of the questionnaire: The questions asked, the language used, and the wording should be appropriate to tap the respondents’ attitude, perceptions and feelings. The choice of words should relate to the educational levels of the respondents. e.g. Some blue-collar workers may not understand the term “organizational structure”. “Working here is a drag” or “she is a compulsive worker” do not interpret in the same way in different cultures. Type and Forms of Questions: Type refers to questions, whether open-ended or closed. The form refers to positive and negative worded questions. 65 Open-Ended versus Closed Questions: Open-ended questions allow the respondents to answer, as they desire. Please look into the following examples. 1. to state five things that are interesting and challenging in the job 2. what the respondents like about their supervisors or their work environment 3. invite comments on the investment portfolio of the company You have to edit open-ended questions have to be edited and categorize for subsequent data analysis. Closed questions have a set of alternatives for the respondents to make choices. It may also ask to rank the first five aspects of the job from the list of 10 or 15 that might be interesting or challenging. Closed questions help the respondents to make quick decisions. It will help the researcher to code the information at ease for analysis later. The alternatives should be mutually exclusive and collectively exhaustive. If not the respondents will get confused. Positive and Negative worded Questions: A good questionnaire should include both positive and negative worded questions. We must avoid double negatives and excessive use of not and only, as they confuse the respondents. Inclusion of some negatively worded questions, will overcome the tendency of the respondents to mechanically circle the points is minimized. Example: It is better to say, “Coming to work is no great fun” than to say, “Not coming to work is greater fun than coming to work”. Similarly, it is better to say, “The rich need no help” than to say, “Only the rich do not need help” Double-Barreled Questions: When the question leads to different possible responses, we call it as double-barreled question. We should avoid such questions, as it will confuse the respondents. Example: “Do you think there is good market for the product and that it will sell well? The answer could be ‘yes’ (there is good market to the product) to the first part and ‘no’ (it will not sell for various other reasons) to the second part. It would be better to ask two questions: 1. Do you think there is good market for the product? And 2. Do think that the product will sell well? The answers could be ‘yes’ for both or ‘no’ for both, or ‘yes’ to the first and ‘no’ to the second and vice versa. Therefore, we should eliminate double-barreled questions. 66 Ambiguous Questions: The respondent might not be sure what exactly the question means. Example: “to what extent would you say you are happy?” It is difficult to decide whether the question refers to their state of feelings at the work place or at home or in general. Different respondents interpret such questions differently and the result would be ambiguous. Ambiguous questions have built-in bias. Recall-Dependent questions: The researcher designs some questions to recall experience from the past. e.g. ‘what was the first job in a particular department and how long’. Answers to such questions will have bias. The respondent will not be able to give correct answers. A better source of obtaining such answers is from the personnel records. Leading Questions: The researcher should not design questions, the way he wants the respondents to respond. Example: ‘Don’t you think that in these days of escalating costs of living, employees should be given pay rises?’ Here, we are pressurizing the respondents to say “yes’. The bias can be less if ask ‘To what extent do you agree that employees do deserve a higher pay rises?’ This question is not suggestive. The response from the respondents shall be: Employees do not deserve a higher pay rise at all: Employees should be definitely be given pay raise : Strongly disagree Strongly agree The in between points can be chosen depending upon the agreement or disagreement. Loaded Questions: When the question is emotionally charged it is called ‘loaded questions’. It is a bias. Example: ‘To what extent do you think that the management is likely being vindictive if Union decides to go on strike?’ The ‘strike’ and ‘vindictive’ are emotionally charged terms, polarizing the management and the unions. If the purpose is two fold, i.e., to find: 1. the extent to which employees are in favor of strike and 2. the extent to which they are afraid to go on strike, as well fears adverse reactions. Social Desirability: If you ask, ‘do you think that older people should be laid off?’ The respondent would give an answer ‘no’ because society will not accept this kind of answer. The answer does not reflect the true feelings of the respondent. Therefore, we need to re-word the question as follows: ‘There are advantages and disadvantages to retain senior citizens in the work force. To what extent do you think companies should continue to keep the elderly on their payroll?’ Certain items that tap social desirability are wantonly included at various points of the questionnaire. An index of each individual is calculated from those responses. This index is then applied to adjust the social desirability biases. 67 Length of Questions: Simple, short questions are preferable to long ones. As rule of thumb, question should not exceed 20 words, or exceed one full line. Sequencing of Questions Funnel approach helps the respondents to progress through the questionnaire with ease and comfort. Questions can be from relatively easy to answer (not require much thinking) to those of progressively more difficult (call for more thought, judgments and decision-making); and general in nature (pertaining to the organization) to those of more specific (incisive questions regarding specific job, department and the like). Positively worded and negatively worded questions taping the same element or dimension of a concept can be placed one after another. For example: 1. I have opportunities to interact with my colleagues during work hours. 2. I have few opportunities to interact with my colleagues during work hours. Classification of Data or Personal Information: The names of the respondents if need be, can be obtained and kept in a private document to ensure the anonymity of the respondent. Questions relating to personal can be placed either at the beginning or at the end as per the choice of the researcher. However, the details of income and sensitive matters can be kept at the end of the questionnaire. A range of response option can be provided instead of asking exact figures as cited below: Age (years) Annual income * Under 20 * Less than Rs.20, 000 * 20-30 * Rs.20, 000 - 30,000 * 31-40 * Rs.30, 001 - 40,000 * 41-50 * Rs.40, 001 - 50,000 We have to collect certain demographic data such as age, sex, educational level, job level, department, and number of years in the organization, even then the theoretical framework does not require. It will help us to describe the sample characteristics. If number of respondents is small, we need not have to collect these details, as they will likely reveal the details of the respondents. To sum up, certain principles of wording must be followed while designing the questionnaire. The questions asked must be appropriate to tap the variable. The language and wording used should be such that, it is meaningful to the employees. The form and type of questions should be geared to minimize the respondent biases. The sequencing of the questions should facilitate the smooth progress of the responses from the start to the finish. The personal data should be gathered with due regard to the sensitivity of the respondents’ feelings, and with respect for privacy. 68 Principles of measurement: We must follow certain principles of measurement, to ensure that the data collected are appropriate to test our hypotheses. Appropriate scales should be used, depending on the required data to be obtained. The interval and ratio scales should be used instead of nominal or ordinal scales. We must assess the goodness of the data through tests of validity and reliability. Validity establishes how well a technique, instrument or process measure a particular concept, reliability indicates how stably, and consistently the instrument taps the variable. GENERAL APPEARANCE OR GET OF THE QUESTIONNAIRE: A good introduction, well-organized instructions, and neat alignment of the questions are important. We now discuss them briefly: A Good Introduction: The identity of the researcher and the purpose of the research are necessary. Researcher must motivate the respondents to respond to the questionnaire and must establish a good rapport with them. Researcher must assure to maintain the confidentiality and ensure less bias on their answers. The researcher at the end of the letter should thank the respondents for having participated in the survey. The following an example of an appropriate introduction. Department of Management Southern Illinois University at Carbondale Carbondale, Illinois 62901 Date Dear Participant, This questionnaire is designed to study aspects of life at work, The information you provide will help us better understand the quality of our work life. Because you are the one who can give us a correct picture of how you experience your work life, I request you to respond to the questions frankly and honestly. Your response will be strictly confidential. Only members of the research team will have access to the information you give. In order to ensure the utmost privacy, we have provided identification for each participant. This number will be used by us only for follow-up procedures. The numbers, names or the completed questionnaires will not be made available to anyone other than the research team. A summary of the results will be mailed to you after the data are analyzed. Thank you very much for your time and co-operation. I greatly appreciate your organization and your help in furthering this research endeavor. Cordially, (Sd) Anita Siglar, Ph.D. Professor 69 Questions must be neatly aligned, with instructions to enable the respondent to complete and answer the questions with least time and effort without straining the eyes. Please refer page No.246 to 248 for example 10.3 to 10.6. Example 10.7: I sincerely appreciate your time and cooperation. Please check to make sure that you have not skipped any questions inadvertently, and drop the questionnaire in the locked box, clearly marked for the purpose at the entrance of your department. Thank you! Review of Questionnaire Design: Questionnaire is useful to collect data from large population in different geographical regions. Obtaining information is easier and coding of responses is fairly easier. All the principles discussed above have to be followed to minimize the bias. The results can be replicated (make an exact copy of) by using well-validated instrument. The questionnaire can be personally administered to respondents, inserted in magazines, periodicals, mailed to respondents, or electronically distributed through e-mail – both through Pre-testing of Structured Questions: It is important that the respondents understand the questions without any ambiguity (uncertain or inexact meaning). Researchers involve small number of respondents to test the appropriateness of the questions before administering them, as well to reduce the biases. We must debrief the results of the pre-test to the participants and must obtain additional information from them on their general reaction to the questions. Electronic Questionnaire Design and Surveys: We can design and administer on-line questionnaire survey when microcomputers are hooked up to computer networks. We can also mail the data discs to the respondents who use their personal computers for responding to the questions. Respondents cannot feel comfortable to respond unless they know how to use the computers. You may refer to table 10.1 for the advantages and disadvantages of interview and questionnaires. PART III: OTHER METHOD OF DATA COLLECTION: We can gather data, without asking questions, by observing the movements, work habits, the statements made, facial expressions of joy anger, and other emotions, and body language of the respondents. Observing the children playing with the toys helps the toy manufacturers to design 70 appropriate toys. The child educators, day care administrators and others involved in the children’s development, design and model ideas based on children’s interests. In the above circumstances, the researcher can be a non-participant observer or participant observer. Non-participant observer: Example: the researcher sits in a corner of a room, observes, records the activities of a manager systematically, and tabulate how the managers typically spend their timings. The researcher will come out with certain findings. However, the observers have to be physically present in the work place and such studies may be time consuming. Participant Observers: He becomes the part of the work team by becoming the employee of the firm. Many anthropological studies (study of human origins) are conducted in this manner by the researcher becoming the part of the alien culture (unfamiliar or unacceptable) for the purpose to study in depth. Structured Vs. Unstructured Observational Studies: Structured observational Studies: Observational studies could be non-participative observer or participative observer. We call it as observational study where the observer has a predetermined set of categories of activity or planned study of a phenomenon. Unstructured Observational Studies: The researcher at the beginning may not have definite ideas of the particular aspects that need focus. He will record everything observed. Such study is unstructured observational study. Once the required information is observed and recorded over a period, we can trace the patterns, and inductive discovery then pave the way for subsequent theory building and hypotheses testing. Observational Studies: Advantages of Observational Studies: 1. It is more reliable and free from respondent bias, generally. 2. It is easier to note the effects of environmental influences on specific outcomes. Example: the weather (hot, cold, and rainy), the sales of a product, traffic patterns, absenteeism etc., can be noted and meaningful patterns might emerge from this type of data. 3. It is easier to observe certain group of individuals. Example: Very young children and busy executives, obtaining information could be very difficult. 71 Disadvantage of Observational Studies: 1. The observer has to physically present for prolonged hours. 2. Collecting data is slow, tedious and expensive. 3. The cognitive thought process cannot be captured. 4. Observers have to be trained to observe what and how, and ways to avoid observer bias. Biases in Observational Studies: Data observed from the researcher’s point of view could be prone to observer biases. There could be recording errors, memory lapses, and errors in interpreting activities, behaviors, events, and normal cues. Inter observer has to be established where more observers are involved. The persons observed may behave differently during the period of study, if the observation is shorter duration. The respondents will be more relaxed as the study progresses and tend to behave normally. Therefore, the researchers discount the data that are recorded in the initial period. Review of the Advantages and Disadvantages of Different Data Collection Methods: Face-to-face interviews: Advantages: 1. It provides rich data and help to explore and understand complex data. 2. Many ideas normally difficult to implement can be discussed and sorted out in face-toface interviews. Disadvantages: 1. Potential for interviewer bias. 2. Expensive if large number of subjects is involved. 3. Adequate training is important where several interviewers are involved. Telephone Interviews: Advantage: 1. Obtain immediate response from the subjects dispersed over various geography regions. 2. It is an efficient way of collecting the data. Disadvantage: 1. Interviewer cannot observe the nonverbal response. 2. The Interviewee may block the call. Personally administering the questionnaires to group of individuals: Advantages: 1. establish rapport with the respondents while introducing the survey 72 2. provide clarifications by the respondents on the spot 3. collect the questionnaires immediately after they are completed 4. one hundred percent response rate. Disadvantages: 1. administering the questionnaires personally difficult, especially the sample is geographically dispersed. 2. the organizations from where data collected should be close to proximity and the groups of respondents should be conveniently assembled in the conference room. Mail questionnaires: Advantages: advantages if the sample is dispersed geographically can be used where telephone interview is expensive Disadvantage: 1. usually have a low response rate 2. the data obtained could be biased as the non respondents may differ with those who had responded Observational Studies: Advantages: 1. helps to understand issues through direct observation. 2. can clarify on certain issues by asking questions Disadvantages: 1. long period of observation will be expensive 2. observer bias may be present in the data Special Issues in Instruments for Cross-Cultural Research: Different languages are spoken in different countries. We must make sure that the translation into the local language matches accurately to the original language. Therefore the translation made from English to Japanese by the local expert has to be translated back to English by a bilinguist to ensure vocabulary equivalence i.e. the words used have a same meaning. The following are the examples in pitfalls in cross-cultural advertising: 1. GM took a step back when it tried to market the NOVA in Central and South America. In Spanish ‘Nova’ means ‘it does not go’. 2. Pepsi’s ‘Come Alive With the Pepsi Generation’, when translated into Chinese, means, “Pepsi brings Your Ancestors Form the Grave’. 3. When American Airlines wanted to advertise its new leather first-class seas to Mexico, its ‘Fly in leather’ campaign would have literally translated to ‘Fly naked’ in Spanish. 4. The ‘Got Milk’ in Spanish would translate to ‘Are you lactating?’ 73 Issues in Data Collection: 1. response equivalence 2. timing of data collection and 3. status of the individual collecting the data. Managerial Advantage: You may collect the data through interviews, questionnaires or observation, by engaging consultants for research purposes. As a manager, you will have the following advantages: You will be able to: 1. phrase unbiased questions to obtain right type of useful responses. 2. decide at what level of sophistication you want data to be collected, based on the complexity and gravity of the situation. 3. understand the dynamics operating in the situation. 4. differentiate between good and bad questions used in surveys with sensitivity to cultural variations, not only scaling but also in developing the entire survey instrument, and in collecting the data. Ethics and the Researcher: 1. treating the information given by the respondent strictly as confidential and should not disclose the identity of the respondent. 2. researcher must explain the nature of the study to the subjects and should not hide information 3. sensitive personal information of the subject can be collected if absolutely required for the project, duly offering specific reasons. 4. the self-respect and self-esteem of the subjects should be never violated. 5. no one should be forced to respond the survey. 6. non-participative-observers should be non-intrusive as possible. 7. in lab studies, the participants should be debriefed with full disclosure of the reason for the experiment after they have participated in the study. 8. the researcher should take personal responsibility for the safety of the subjects and the subjects should never be exposed to physical or mental harm. 9. there should be any misrepresentation or distortion in reporting the data collected during the study. Ethical Behaviors of Respondents: 1. once the subjects exercised the choice to participate in a study, they should fully cooperate. 2. the respondents or subjects have an obligation to be truthful and honest in the responses. they should not misrepresent or give information knowing it to be untrue. ----74 DATA ANALYSIS AND INTERPRETATION We analyze the data collected from a representative sample of the population to test the research hypotheses. The data collected should be good and has assured quality for further analysis. Coding can be done for the data collected viz. age, education, job level, sex, work shift, employment status etc. The following four steps play vital role in data analysis and interpretation. 1. getting data ready for analysis 2. getting a feel for the data 3. testing the goodness of the data 4. testing the hypotheses GETTING DATA READY FOR ANALYSIS Editing Data The interviewer, or observer or researcher might have noted the responses to the open-ended questions of interviews and questionnaires, or unstructured observations, in a hurry. They must edit the data on the same day. It must be clearly deciphered, and coded in its entirety. The edited data must be identifiable by a different colour of pencil or ink, so that original information is still available in case of further doubts later. We have to check the data from the questionnaires for incompleteness and inconsistencies and edit them logically. For example, a respondent might have inadvertently omitted to fill the column against whether married. However, might have filled the details relating to the number of years married, and number and ages of children. These details confirm her marital status but possibly she did not want to disclose, because she is a widow or has been divorced, or for some other reason. We will introduce a bias in the data and edit the data to read as ‘yes’. This case is clear-cut for editing. We have to follow-up with the respective respondents and correct / edit the data. Certain other omissions could be left unnoticed or can not be rectified. The researcher may not have any control over other biases affecting the goodness of the data. The validity and replicability of the study could thus be impaired. Much of the telephone and electronically administered questionnaires will automatically edit, while the respondent is answering the questionnaires. Handling Blank Responses The respondents leave the answers blank, may be, for the following reasons: 1. did not understand the questions 2. not willing to answer the questions 3. simply indifferent to respond to the entire questionnaire If a substantial number of questions were not answered (say, 25% of the questionnaires), you should not include the same for data analysis. You should mention the number of such unused responses, due to excessive missing data, in your final report. If two or three items left blank in 30 questions, you may decide how these blank responses can be handled. 75 There are several ways to handle blank responses as follows: 1. give a mid point in a scale as the value 2. ignore the particular item during the analysis 9reduce the sample size for the given variable) 3. assign mean vale for all those responded to the particular item 4. assign mean value to all other questions measuring the variable of this particular item 5. a random number with in the range for that scale Please note SPSS (software program) uses linear interpolation from adjacent points as also a linear trend to replace missing data. If many of the respondents have answered ‘don’t know’ to a particular item or items, the questions might not have been clear / understood or some other aspects could have prevented them from answering. In this circumstance, we must do further investigations. Coding We have to code the responses. Scanner sheets make it easier to enter the responses directly into the computer without manual keying in. If you can’t use a scanner sheet for any reason, use a coding sheet to transfer the data from the answer sheet. Please refer to table 12.1, relating to coding the Serakan data – the answer to Exercise 10.4 in chapter 10.4 in respect of the questionnaire design exercise to test the job involvement and job satisfaction hypotheses in the above referred company. Coding can become simple. Numbers given in the boxes instead of giving a box to mark the appropriate box help transfer to the coded sheet easier. Coding can be done by using the actual number circled by the respondents. Keying the data directly to the computer from the answer sheet may result possible errors and omissions of items. Therefore, we must use a coding sheet. Human error can occur in coding. We must check at least, 10% of the coded questionnaires for accuracy by following the systematic sampling procedure. If you find many errors in the sample, check all the coded questionnaires. Categorisation 1. categorise the variables – group the items measuring the concept 2. reverse the negatively worded questions so that all the answers will be in the same direction (strongly agree means strongly disagree). Questions 16 to 21 in the Serakan Co have to be recoded, i.e. the scores of 7 read as 1; 6 as 2; 5 as 3; 3 as 5; 2 as 6; and 1 as 7. Questions measuring a concept are not contiguous (sharing a border) but may be scattered over various parts of the questionnaire. You must take care that to include all the items without any omission or wrong inclusion. Entering the Data If data are not collected in the scanner sheets, the raw data can be manually keyed into the computer through any software program. SPSS Data Editor, looks like spread sheet, can enter, edit and view the contents of the data file. Each row of the editor represents a case, and each 76 column represents a variable All missing variable will appear with a period 9dot) in the cell. You can add, change, or delete values after entering the data, if required. Once the missing vales, the recodes, and computing of the new variables are taken care of, the data is ready for analysis. DATA ANALYSIS SPSS Version 11.0 and Excel programs can be used to interpret the results of the analysis or use any other software to produce similar results and interpreted in the same manner. Basic Objectives in Data analysis We have three objectives in data analysis: 1. feel for the data 2. testing the goodness of the data 3. hypotheses testing Feel for the Data The following statistics give the feel for the data, irrespective of whether or not the hypotheses are directly related to the analysis. 1. the frequency distributions (central tendency and the dispersions), for the demographic variables (the study of the structure of human populations using records of number of births, deaths, instances of disease etc.) 2. the mean, standard deviation, range, variance on the other dependent and independent variables, and 3. an inter-correlation matrix of the variables These statistics give a feel for the data. The examination of the central tendency, and how clustered or dispersed the variables are, gives a good idea of how well the questions were framed for taping the concept. The correlation matrix will give an indication of how closely related or unrelated the variables under investigations are. If the correlation between two variables are high (over .75), we would surprise whether they are two different concepts or measuring the same concept. We will begin to wonder whether we have measured the concepts validity and reliability. We do a detailed analysis after this initial feel, to test the goodness of the data. Testing the Goodness of Data The reliability of a measure is established by testing for both consistency and stability. Consistency indicates how well the items measuring the concept hang together as a concept hang together as a set. Cronbach’s alpha is a reliability coefficient, how well the items in a set are positively correlated to one another. Cronbach’s alpha is computed in terms of the average inter-correlations among the items measuring the concept. The closer Cronbach’s alpha is to 1, the higher the internal consistency reliability. 77 We use another measure of consistency in specific situations is the split-half reliability coefficient. This reflects the correlations between two halves of a set of items. The coefficients obtained will vary depending on how the scale is split. We obtain split-half reliability to test for consistency when more than one scale, dimensions, or factor, is assessed. We split the items across each of the dimensions or factors based on some predetermined logic. Cronbach’s alpha is an adequate test of internal consistency reliability for almost every case. The stability measures can be assessed through parallel form reliability (when a high correlation between two similar forms of a measure is obtained) and test-retest reliability (computing the correlation between the same tests administered at two different times). Validity The results of factor analysis (a multivariate technique) will confirm whether the dimensions theorized, emerge or tapped by the items in the measure. Criterion-related validity can be established by testing for the power of the measure to differentiate individuals who are known to be different. Convergent validity can be established when there is high degree correlation between two different sources responding to the same measure (e.g. both supervisors and subordinates respond similarly perceived reward system measure administered to them). Discriminant validity can be established when two distinctly different concepts are not correlated to each other (e.g. courage and honesty; leadership and motivation; attitudes and behavior). Hypothesis Testing / Data Analysis and Interpretations The out of the range / missing responses, etc. are cleaned-up and the goodness of the measures is established, the data will be ready for analysis. After a brief description of the background of the company we took for research and the sample, we will discuss the analysis done to obtain the feel for the data, establish reliability, and test each hypothesis as well how well the results are interpreted. We will now examine the results of analysis of data obtained from a company, and the interpretations. Please refer to pages from 308 ‘Uma Sekaran’ Please refer to pages 138 to 149 in respect of Measures of relationship, simple regression analysis and association in case of attributes. 78 FACTOR ANALYSIS We use factor analysis to study complex product or service to identify the major characteristics considered important by the consumer of the product or service. Example: Researchers for an automobile company may ask large sample of potential buyers to report (7 or 10 or 11-point scale), the extent of their agreement or disagreement with the following factors: 1. The side profile of the car should be sleek. 2. A car’s breaks are its most critical part. 3. Identify safety, exterior styling, interior rooming or economy of operations by potential customers. Note: Researchers use interval scale or continuous scale to measure the variables in respect of the above example. Factor analysis guide to design the product to meet the needs or expectations of the prospective consumer or identify the themes that potential customers consider important. What Factor Analysis does Using the data from the large sample, factor analysis applies advanced form of correlation analysis to the responses to a number of statements. The purpose of analysis is to determine if the responses to several of the statements are highly correlated, i.e. statements measure some factor commonality among them. Example: Consider the following statement: 1. A car’s breaks are its most critical part. 2. I want my next car should be equipped with an ‘air bag’ 3. A collapsible steering column should be a standard one in all cars. The above set of statements indicates an underlying concern with the factor of safety. Factor analysis involves many statements. The statements in any one set are highly correlated but are not highly correlated with the statement in any other set. Types of variables used in Factor Analysis Factor analysis can only applied to: 1. Continuous variables 2. Interval scaled variables Factor Analysis identifies interdependencies among variables 1. Factor analysis uses more than one variable to identify a class or category that is important from a marketing standpoint. 79 2. Factor analysis identifies two or more questions that result in sets of responses that are highly correlated. An example of Factor Analysis application, in respect of a compact car: 1. 2. 3. 4. 5. A car’s breaks are its most critical part. I want my next car should be equipped with an ‘air bag’ A collapsible steering column should be a standard one in all cars. Four adults should be able to comfortably sit in a compact car. Mileage in a compact car should be, at least, 18 km. per liter of petrol. Three hundred individuals gave their responses to 100 statements each on a 7-point scale in respect of the above-mentioned example. There were 100 such distributions, one for each of the 100 statements. The researcher applies factor analysis to the data to identify the major characteristics that potential buyers of the compact cars consider important. In this regard, the following description treats five factor analysis topics: 1. 2. 3. 4. 5. three important measures the role of correlation the identification of factors the output of factor analysis evaluating how well the fit Three Important measures 1. The variance 2. Standardised Scores of an individuals responses 3. The role of correlation Variance: A factor analysis like regression analysis tries to ‘best fit” factors to a scattered diagram of the data to show that factors explain the variance associated with the response to each statement. Regression equation fitted to a scattered diagram of responses to variable ‘y’ and ‘x’ helps to explain the variance observed in the responses to variable y. A user of regression analysis would like to explain 100% of the variance, in a dependent variable – i.e. get an R2 = 1.00 -, the user of factor analysis would also like to explain 100% of variance associated with each statement used in the study. Standardised Scores of Individual Responses: Answers to some questions may be recorded on a 7-point scale and some other questions on a 10-point scale. For the purpose of comparison, the responses to the questions on different scales have to be standardised as explained below: 80 It is possible to calculate the mean and the standard deviation of all the responses to each statement. Similarly, an individual’s actual response to a statement can be standardised by using the following relationship. Individual’s actual Mean of all 300 response to the responses to the the statement statement Individual’s standardised Score on the statement = Standard deviation of all the 300 responses to the statement The individual standardised score is the actual response measured in terms of number of standard deviations (+ or -), it lies away from the mean. Therefore, each standardised score is likely to be a value somewhere in the range of +3 and -3 with +3.00 typically being equated to the ‘agree very strongly’ and -3.00 typically being equated to the ‘disagree very strongly’ response. The Role of Correlation The role of correlation in factor analysis can be explained by using fewer than 100 statements. We assume using six statements as mentioned in page No.2, and calculate the correlation coefficients for all the possible pairs of statements. To illustrate the role of correlation coefficient factor analysis, we assume two factors exist in the set of data. Please refer to table 17-2 matrix of correlation coefficients between pairs of statement for example of two factors and the following explanation: The correlation coefficient associated with the response to statements 1 and 2 shows perfect correlation existed between the two statements. There is also very high correlation resulted from the responses to statement 2 and 3. The responses to statement 4, 5 and 6 are also highly correlated with each other. However, the responses to statements 1, 2 and 3 are highly correlated, but are completely uncorrelated with responses to statement 4, 5 and 6. Similarly, the responses to statements 4, 5 and 6 1, 2 and 3 are highly correlated, but are completely uncorrelated with responses to statement 1, 2 and 3. From this findings, the researchers have evidence to suggest that two factors exists in the data – one factor associated with statement 1,2 , and 3 and another factor with statement 4,5, and 6 Basic concept in the role of correlation 1. The statement in any set need to be highly correlated with each other say r = 0.7 or larger. They need not have to be perfectly correlated. 2. As different sets of statements are relatively uncorrelated with each other, a separate or direct factor is associated with each set. The identification of factors Factors are linear equations of variables (i.e. the statements), measured during the course of the study. Fig. 17-3 shows a scattered diagram of the standardised scores on two factor analysis 81 variables X1 and X2. The figure also shows two factors fitted to the data. We can write the equation for the said two factors as follows: First factor equation Second factor equation : F1 = 0.6 X1 + 0.4 X2 : F2 = 0.4 X1 + 0.6 X2 Thus, each factor is a weighted, linear combination of the two variables being analyzed. For example, where four factors are involved the terms of variables can be expressed as X1, X2, X 3, and X 4. The important measures used in factor analysis are – the variance associated with the standardised responses to each statement in the study. Factor analysis selects one factor at a time using procedures that ‘best fir’ each other to the data. Each additional factor explains less of the variance than the first factor or any other factors identified previously. Each factor selected after the first factor must be uncorrelated with factors already selected. This process continues until the procedure cannot find additional factors that significantly reduce the unexplained variance in the standard scores. Factor loadings Please refer to table 17-3. The 18 numbers located in the six rows and three columns are called factor loadings, one of the three useful output obtained from a factor analysis. Please refer to Fig 17-4a for the statements mentioned in table 17-3 for the concept of high correlation. Factor 1 is highly correlated with the responses to statement X1 (0.84 correlations) and with responses to statement 2 (0.84 correlation). Please refer to Fig 17-4b for the statements mentioned in table 17-3 for the concept of how uncorrelated. Here, the statements 1 and 2 are not highly correlated (012 and 0.18 respectively) with factor 2 as illustrated in Fig. 17-4b. Thus, a factor loading is a measure of how well the factor fits the standardised response to a statement. Naming Factors and measuring their Importance From table 17-3: Factor 1 (F1) is a good fit on the data from statement 1, 2, and 3 but a poor fit on other statements. Statement 1, 2, and 3 probably measuring the same basic attitude or value system, confirms that a factor exists. “Economy of operations” was the factor that tied these statements together in the minds of the respondents. Researchers now wanted to know whether 300 respondents participated in this study mostly agreed with or disagreed with the statement 1, 2 and 3. They found the means of these responses were +0.97, +1.32, and +1.18 respectively for statement 1, 2, and 3, indicating that most of the respondents agreed with the three statements, the researchers concluded that factor ‘economy of operation” was important in the minds of the compact car buyers. 82 Factor 2 (F2), is a good fit on the statements 4 and 5, but a poor fit on other statements. This factor is different from statements 1, 2, 3 and 6. Researchers concluded that factor “interior roominess” was important. Factor 3 is a good fit on statement 6 relating to “safety”. As there were two and one statement for factor 2 and 3, the researchers were less confident of identification of F2 an F3. The researchers concluded “interior roominess” is an important factor with statements 4 and 5 (with the means of +0.91 and +1.22 respectively). As the mean was + 0.07 in respect of statement 6, the researchers were unable to conclude, that “Safety” was considered important. Evaluating How well the data Fits (Second factor) Communalities indicate the proportion of the variance in the responses to the statement, which is explained by the three identified factors. For statement 5, three factors explain 0.89 (89%) but only 0.54 (54%) variance for statement 3. the table 17-3 shows that three factors explain 75% or more of the variance associated with statements 1,2,4,5, and 6, but only half of statement 3’s variance. Researchers use communalities to find out ‘how well the factors fit the data’. As three factors account for most of the variance stated with each of the six statements, the three factors fit the data quite well. Eigen Value (third factor) Eigen value helps to find out ‘how well a factor fits the data from all of the respondents on all the statement’. There is an eigen value associated with each of the factors. When a factor’s eigen value is divided by the number of statements used in factor analysis, the resulting figure is the proportion of the variance in the entire set of standardised response scores, which is explained by the factor. Example: Factor F1, explains 0.3226 (or 32.26%) of the variance of the standardised scores from all of the respondents on all six statements. By adding the variance of the standardised scores for all the three factors, the variance for the entire set of response data is 77.07% (0.3226 + 0.3090 + 0.1391 = 0.7707 (or 77.07%). This figure can be used as a measure ‘how well, overall, identified factors fit the data’. In general, a factor analysis that accounts for 60-70% or more of the total variance can be considered a good fit to the data. 83 Usefulness of Factor analysis Users of the product or services have difficulty in identifying the characteristics that are important to them. The advantage of using Factor Analysis helps the researchers to identify the important characteristics of products or services perfectly that are so complex. Problems in using Factor analysis 1. A factor analysis is of little use if the a. appropriate variables have not been measured, or b. if the measurements are inaccurate, or c. if the relationship in the data are nonlinear 2. Deciding how many identified factors one should use in factor analysis. In example referred in page 5 of this note, the third factor explains an additional 13.91% of the total variance. The addition of this variance substantially increases the variance from 63.16% to 77.07%. The increase is almost one fourth. The said increase is reasonable. We may not use the third factor, if the increase explains only 2 or 3 %. We may not. 3. The third difficulty is to identifying and naming of the factors. For example ‘economy of operation’ in a compact car, we may not know exactly the measure of this factor. 84 CLUSTER ANALYSIS We can find the application of cluster analysis in market segment studies. Researchers use cluster analysis to segment the market based on several attitudes or variables. Cluster analysis is different from identifying different market segments based on one variable only (e.g. heavy users, average users, or prefer brand A or prefer brand B). We can study in cluster analysis the different segments that exist in the total market for any given product (e.g. sports goods). The analysis identifies clusters of respondents who have given the same answers to a certain combination of questions. We use interval scale or continuous scale to measure the variable. Example: A large sample of users is asked to report their attitudes regarding their preferences for indoor or outdoor sports, their preference for rugged or easy sporting activities. Here, we use cluster analysis on the above-mentioned data to see whether total market consists of a number of different segments. What Cluster analysis does Cluster analysis identifies different groups – e.g. the respondents in one cluster are similar to each other but different from the respondents in any other clusters. Cluster analysis is applied to data consisting of many variables collected from a large sample of respondents. The cluster analysis set the procedures, search through the data and identifies identical or similar answers to certain combination of the questions. These respondents are formed into one character. Then search through the data looking for a second set of respondents, who have given similar answers to some other combination of questions. Thus, the second cluster respondents are similar but quite different from the First cluster. Similarly, we can identify third cluster, which is different from the first two clusters. This procedure has to be continued until all of the clusters have been identified. Cluster Analysis identifies Interdependencies among variables: Some of the Multivariate methods – cross-tabulation, LDA (linear Description Analysis) and AID (Automatic interaction Deduction) are concerned with a single variable to identify a class or a category into which a respondent belonged. Multiple-variable classification is used whenever it is useful to marketing decision than singlevariable classification. Cluster analysis is concerned with interdependencies among a number of variables measured in the study, especially with different sub sets of respondents. This will help researchers to understand when and how the cluster analysis can be better used. 85 An Example of Cluster analysis A large sample of sports-active individuals was asked to respond following in a 10- point scale. 1. Their preference for indoor vs. outdoor sports (variable X1). 2. Their preference for rugged and heavy Vs. Easy and light activities (variable X2). Please refer to figure 17-1 for plot of data from Sporting Activities and Interests Study. Each dot in Fig.17-1 represent the score on X1 (vertically) and X2 (horizontally). Researchers can use this figure 17-1 to look for patterns of responses in the geometrical space defined by variables X1 and X2. Researchers will look whether the number of respondents answered two or more questions in the same or similar way. There were 12 respondents. Respondent No.6 reported a score of eight on variable X1 and two on variable X2. The scores of other respondents had also been plotted in the Fig.17-1. There are three clusters with respondents 9, 11, and 12; 6, 7, 8 and 10; and 2, 3, and 4 respectively. Market for sporting equipment can be segmented based on: a. indoor Vs. outdoor b. rugged Vs. light activities Respondents 9, 11, and 12 are interested in hunting, mountain claiming and motor cycle racing. Respondents 6, 7, 8, and 10 are interested in hiking, fishing and camping. Respondent 2, 3, and 4 enjoy in racquetball, indoor tennis and gymnastics. They are interested in indoor activities which are not too easy, but also not too rugged. Respondent 1 is a bowling enthusiast who preferred easy indoor activity and Respondent 5 prefer rugged activities either indoor or outdoor. The above-referred example involves only two variables. Another variable X3 is included; it can be presented in three-dimensional space. If there are four or more variables, the data can be laid in “n-dimensional space” where ‘n’ represents the number of variables in the study. Researchers will not be able to visualize if the study involves more than two or more variables. Under such circumstances, we can use the following procedure to analyze the data to determine whether the clusters do exist. 1. Developing measures to identify similar respondents and 2. Developing procedure for grouping similar respondents Distance: A Commonly used Similarity Measure Researchers must use some measure to identify the similarity between two respondents. If there is no ideal way to measure the similarity, ‘distance’ between two respondents will be used as a 86 measure of similarity. Please refer to Fig.17-1, where the distance separating respondents 7 and 8 is only one unit of variable X2. The distance between respondents 7 and 9 is seven units of variable X2. The distance between respondents 6 and 7 are small they can be considered more similar than the respondents 7 and 9 as not very similar, because of large distance. When the distance is small, then group them into one cluster. Introducing a New Variable Symbol So far, we have identified a variable with a single subscript viz. X1, X2, X3, ….. We now need to introduce a new variable symbol with two subscripts X11, X15, … and X21, X25, … The first subscript identifies the variable number (the score on the variable viz. the score) and the second subscript identifies a specific respondent. X11 means variable X1 and respondent No.1. The symbols used to identify the reported scores on variables 1 an2 by respondents 1, 5, and 6 are given below: Variable 1 symbol Score X11 2 X15 5 X16 8 Respondent No. 1 5 6 Variable 2 symbol Score X21 1 X25 8 X26 2 The ‘Distance’ formula We commonly use Euclidean geometry to measure the distance in cluster analysis. The distance between respondents 1 and 5 will be identified by a symbol D15. The distance between 1 and 5 in a study where variables X1, X2, X3, …… Xn is calculated with the following formula: D15= (X11 - X15)2 + (X21-X25)2 + (X31-X35)2 + … + (Xn1-Xn5)2 The second subscripts associated with each variable (1 and 5) identify that the formula is being applied to data from respondents 1 and 5. X21 represents the data on variable 2, obtained from respondent 1 and X35 represents the data on variable 3, obtained from respondent 5. The distance between respondents 1 and 5 is: D15= (X11 - X15)2 + (X21-X25)2 D15= (2 - 5)2 + (1-8)2 D15= 9 + 49 = 7.6 Please refer to table 17.1. Row 1 in column 5 the distance is 7.6 units between respondent 1 and 5. similarly, the distance between 2 and 9 are separated by a distance of 8.1 units as shown in Row 2 in column 9. 87 The Single linkage rule: The respondent will be placed into a group if the distance between the particular respondent and any other single respondent already in the group is smaller than the pre-established minimum distance (MD). The single linkage rule considers the following to form clusters: 1. 2. 3. 4. Pre-establish minimum distance (MD) Form clusters (groups) who are very close by Then include the respondents moderately close by Include the respondents within MD The procedures to form clusters are illustrated in Fig.17.2 for the data shown in Fig.17-1 and table 17-1. Usefulness of Cluster Analysis 1. Used to identify different segments in a market based on a number of attitudinal or behavioral variables as discussed in the above example. 2. Used to study different types of perfume users, different types of husbands, different types of beer drinkers, and many other potential market segmentations. Problem in using Cluster analysis Researchers can encounter certain problems when using the cluster analysis. Careful thought should be given to the following: 1. 2. 3. 4. 5. The variable to be measured The similarity measure to be used Grouping procedure to be used The selection of MD value How good the Cluster Analysis is? If minimum distance (MD) value is too large, the respondents who are not having similarity will be included in the cluster. If minimum distance (MD) value is too small, many respondents will not be included in any of the clusters. We cannot perform any statistical test as to what is the ideal number of clusters. 88 THE RESEARCH REPORT THE RESEARCH PROPOSAL There must be an agreement to the following, before undertaking a research study, between the researcher and the person who gives official permission to the study: 1. 2. 3. 4. as to the problem to be investigated the methodology to be used the duration of the study, and its cost This kind of agreement is required to ensure that there are no misunderstandings or frustrations later for both parties. The researcher / investigator carefully organize his proposal and submit the research proposal duly taking into account of the following factors: 1. 2. 3. 4. the broad goals of the study the specific problem to be investigated details of the procedures to be followed the research design offering details on: a. The sampling design b. Data collection methods c. Data analysis 5. time frame of the study, and when the written report will be handed over to the sponsor 6. the budget - detail the costs with reference to specific items of expenditure. The sponsor might seek some clarifications on some points, require some modifications in certain respects, or may accept in toto. Please refer to a simple research proposal to the study retention of new employees as presented in Model 13.1. The report The results of the study have to be effectively communicated to the sponsor. Writing the report consciously, convincingly, and with clarity is important, than conducting a perfect research. The oral presentation is also equally important. The written report and the oral presentation depend upon the purpose of the study and target audience. The written report The manager weighs the pros and cons of the research report to implement the acceptable recommendations. He compares the existing state of affairs and the desired state to bridge 89 the gaps if any. Therefore, the written report has to focus on the issues discussed below: The written report and its purpose The forms of report vary according to the purpose and situation. Specific areas of interest: The report can provide desired information to the manager in a brief format and can focus narrowly. Please refer to example 13.1. ‘Sell an idea’ to the management: You need to present all the relevant information supporting with necessary data. The report should be detailed and convincing as to how the proposed idea is an improvement and how to adopt. This strategy will persuade the reader to ‘buy into the idea’. Please refer to example 13.2. Several alternative solutions or recommendations: The researcher provides several alternative solutions or recommendations to the manager and the manager chooses the alternative and makes a final decision. The researcher has to provide a detailed report in respect of the following: 1. 2. 3. 4. 5. 6. 7. 8. more detailed report surveying the past studies the methodology used for the present study different perspectives generated from interviews current data analysis and alternative solutions based on the conclusions drawn there from discussion on how each alternative helps to improve the problem situation advantage and disadvantage of each of the proposed solutions and cost-benefit analysis in terms of rupee / or other resources Please refer to example 13.3 to the above. Identify the problem and provide the final solution: The sponsor calls the researcher to study a situation, determine the nature of the problem and offer a report of the findings and recommendations. Such report must be comprehensive after a detailed study. Findings of a basic study: It is a very scholarly publications published in academic journals. Such report must be comprehensive after a detailed study. The written report and its audience You must organize the report, its length, focus on details, data presentation and illustrations and focus them to the intended (having in mind) audience. Some reports may be long and detailed and some short or brief and specific. An executive summary in less than three pages at the beginning will help the busy executives to grasp (understand) the essentials (important details) of the report findings and look into pages for information of special interest. A good report will meet the purpose of the person who is interested in the report. 90 Some managers will feel distraction in seeing the tables and prefer graphs and charts. Some others may want only ‘facts and figures’. Both figures and tables are visual forms and will have presence in the report. Depending upon the idiosyncrasies (a person’s particular way of behaving or thinking) of the person who is going to use the report, the researcher can highlight the charts or figures in the report or find a place in the appendix. Where different executives are going to handle the report, they should know where they could find the information. For example, the market share of the company can be mentioned in the text. Further, we can present a pie chart and the raw data in a tabular form. Findings of the study, sometimes, unpalatable (not pleasant to be acceptable), because of outmoded policies of the company or is highly bureaucratic). In such situations, you must present the data supporting the facts. In addition, must say that these policies were appropriate at the time of introduction. You must highlight that the present system is receptive of changes and hence, requires modifications / changes considering the present requirement. When the system has an ineffective top-down approach, we can follow the similar strategy cited in this paragraph. Tact and diplomacy combined with honesty and objectivity are important in report writing. The task of the internal research team will be difficult for the situation referred in the previous paragraph, than the external research team. Characteristics of well-written reports The important features of well-written reports are: 1. 2. 3. 4. 5. 6. 7. clarity conciseness ( giving a lot of information with clarity) coherence the right emphasis on important aspects meaningful organization of paragraphs smooth transition from one topic to the next apt choice of words Avoid technical or statistical jargons, unless it is required. Check for grammatical and spelling errors. Explain the assumptions, if any, made in the report. Ensure good appearance of the report. You must give appropriate headings and sub-headings and organize the report in a logical manner. This will help the reader to follow the transition (change of topics) easily. Use double space typed report with a wide margins on all sides, help the reader to make notes/comments while reading the report. Contents of the research report A well-conducted study loses all its value when it is not properly presented. The following discussions will help enhance the writing skills: 91 A research report should bear a title that can clearly indicate what the study is about. It should contain a table of contents, the research proposal, a copy of authorization to conduct the study and an executive summary (in the case of applied research) or a synopsis (in the case of basic research) You should give an introduction detailing the purpose of the study with some related background, state the problem studied and give an idea what the reader can expect in the report. The body of the report shall contain the framework of the study, hypotheses, if any, sampling design, data collection methods, analysis of data, and the results obtained. The final part of the report shall deal with presentation of findings and draw conclusions. Recommendations would contain cost-benefit analysis from which we can come to know the net advantage of each implementation of the recommendations. The thoroughness of the report should instill confidence in the minds of the reader. At the same time, you should highlight limitations of the study, if any (example: in sampling, data collection, etc.). Provide scientific authenticity to the report. Write with personal touch, wherever required. Provide a good rationale (based on reason or logic) for the study. Present the problem studied clearly. Present the results of data analysis fully and adequately and interpret (clarify) the data enabling the reader to understand easily. The conclusion drawn from the findings should indicate a clear solution to the problem We can organise the report in parts or chapters and to be tailor made to meet the needs of the situation. Some of the essential characteristics of a good report are: ‘good, crisp, and clear writings, figures, charts and tables that succinctly (briefly and clearly expressed) support or highlight the salient issues, and attractive packaging” The writing style should be simple, interesting, precise and comprehensible. Unbiased and objective presentation of the findings of the study lends credibility to the research work. Tact and diplomacy are required and do not offend the sponsor. You should tailor the format and style of report to requirement of the audience and must meet the purpose of the study. End the report with a summary. Acknowledge the help received from various individuals and sources. Cite references in the report. Attach appendices, if any, to the report. 92 INTEGRAL PARTS OF THE REPORT The title page The title of the report succinctly (briefly and clearly expressed) indicate what the study is about. Examples of some good report titles are: 1. A study of Customer Satisfaction with the Pizza Hut at Chennai, Tamilnadu 2. Factors Influencing the Burnout of Nurses in Apollo Hospital 3. Antecedents and Consequences of White-Collar Employees Resistance to Mechanisation in Service Industries 4. Factors Affecting Upward Mobility of Women in Accounting Firms 5. A Study of portfolio Balancing and Risk management in Investment Firms The first two projects relate to applied research and the last three refers to basic research. The title page will include the name of the sponsor of the study, the names of the researchers and their affiliations, and the date of the final report. Table of contents It contains the important headings and sub-headings in the report. Give a separate list of tables and figures in the table of contents. The research proposal and the authorisation letter At the beginning of the report, attach the copy of the sponsor’s authorization letter along with the research proposal. From this, the reader will come to know that the study had the full backing of the sponsor. The executive summary or synopsis The executive summary (synopsis) will be brief and normally three pages or less. It provides a brief account of the research study, providing an overview, and highlights the following important information related to the research report: 1. 2. 3. 4. 5. the problem statement sampling design data collection method used results of data analysis findings and recommendations, with suggestion for implementation. An example of the study of customer satisfaction with the Pizza Hut in Sunshine city can be seen in example 13.4 (enclosed). --- 93 RESEARCH METHODS IN BUSINESS 16 mark questions 1. Are scientific investigation and managerial decision making integral part of decision making? 2. How will you apply the various hall marks of scientific research? 3. Explain the hypothetico-deductive method. 4. How literature survey is useful in a research study? 5. Explain the research process. 6. How will you conduct literature survey? 7. Developing a good theoretical framework is central to examine the problem under investigation. Discuss. 8. Explain experimental designs with suitable examples. 9. How will you control the variables in an experiment? Explain with examples. 10. Enumerate the factors that are affecting the internal validity and external validity 11. What are the types of experimental designs and internal validity? 12. Explain Solomon Four-group Design. 13. Explain the steps in sampling design and the characteristics of good sampling design. 14. What are the different types of sampling methods? 15. Explain various methods of data collection. 16. Enumerate questioning techniques with suitable examples. 17. What are the various types and forms of questions used in a research? 18. Explain the advantages and disadvantages of different data collection methods: 19. What are the kinds of ethics a researcher has to follow in a research process? 20. What are the steps that play a vital role in data analysis and interpretations? 21. Why do you use factor analysis to study complex product or service? Explain with an example. 22. What is the role played by correlation in factor analysis? 23. What cluster analysis does? Explain with suitable examples. 24. How will you write a research report? What factors will you take into account while writing a research report? 25. What are the characteristics and contents of a well written report? 26. What are the integral parts of a research report? 27. What factors would you take into account before accepting to do a research? 94 RESEARCH METHODS IN BUSINESS 2 mark questions 1. Define applied and basic research. 2. Briefly explain replicability. 3. What do you mean by generalisability? 4. List few obstacles to conduct research. 5. Differentiate deduction and induction. 6. Briefly explain the broad problem area with an example. 7. Define dependant and independent variables. Give examples. 8. How moderating variable affect the dependant variable? 9. Why you need to manipulate the independent variable? 10. What do you understand from the term ’theoretical framework’? 11. Differentiate lab experiment and field experiment. 12. What are the advantages and disadvantages in randomization? 13. Explain briefly the internal and external validity. 14. How confounding variables can be randomly distributed? 15. Give at least two unethical practices that can be avoided. 16. Give a brief on census sample survey. 17. What do you mean by convenience sampling and quota sampling? 18. Explain briefly, how will you collect data through observation method? 19. How will you analyze the data? 20. What are the good qualities of research? 21. What are the problems encountered by researchers in India? 22. Define primary and secondary data. 23. What do you understand from Unobtrusive or Trace Measures in respect of data collection methods? 24. Why you need to take care on language and wording the questionnaires? 25. Differentiate open ended and close ended questionnaires. 26. How will you deal blank responses in questionnaires? 27. Give the uses of factor analysis. 28. How single linkage rule forms clusters? 29. What are the uses and problems in cluster analysis? 95