Research methodology for behavioral research

Research Methodologyfor Behavioral Research


• Objective– To introduce RM and multivariate data analysis

for behavioral research to new Ph.D. students– To provide new Ph.D. students with hand on

experience on structural equation modeling by LISREL.

• Topics– Overview– Conceptualization


– Measurement (Reliability and Validity) and Scale Development and Validation

– Research Design: Survey Research– Research Design: Experiments– Multivariate Data Analysis– Structural Equation Modeling by LISREL– Conclusion

Overview

Research MethodologyOverview

• Types of Research– Classification based on Data Collection Method

• Primary vs. Secondary Research– Classification based on the Treatment of Data

Collected• Qualitative vs. Quantitative Research

– Classification based on the Objective of Research• Exploratory vs. Descriptive vs. Conclusive

Research

Research

Primary researchSecondary research(Literature review)

Qualitative Research Quantitative Research

Exploratory Research Descriptive ResearchConclusive/Causal

Research


• Steps in Research Methodology– Step 1: Problem formulation– Step 2: Hypothesis formulation– Step 3: Research design– Step 4: Sample and sampling– Step 5: Data collection– Step 6: Data analysis– Step 7: Manuscript Writing


• Important Concepts– Concepts– Constructs– Variables– Relationships among concepts, constructs, and

variables– The role of concepts, constructs, and variables

in research methodology


• Concepts– Concepts are abstractions from partciculars.– Concepts have constitutive definitions. So

concepts are rich in meaning but cannot be measured directly.

– Many things we want to study in behavioral research are concepts, e.g., quality, satisfaction, attitude, leadership.

– In research methodology, concepts are used in the step of problem and hypothesis formulation.


• Constructs– Constructs are concepts that are measurable.– Constructs are measurable because they have

additional definitions, operational definitions.– Operationalization of concepts into constructs

concern with the concept of validity and reliability.

– After operationalization, each concept becomes a construct. In the measurement instrument (e.g., questionnaire), each construct becomes a measurement scale.


– A measurement scale can be a single-item or multiple-item scale.

– In research methodology, constructs are used in the step of designing the measurement instrument (operationalization of concepts).

• Variables (or Observed Variables)– After using the measurement scales to collect

the responses from the subjects, each response to each measurement of the scales is then coded into “number”.


– After “number” being assigned to each measurement item of the scales, each item becomes an observed variable.

• Relationships among Concepts, Constructs, and Variables– 1 concept 1 construct 1 scale 1 or

more observed variables


• Step 1: Problem Formulation– Qualitative research (e.g., focus group, in-depth

interviews) can be used to help formulate problems.

– Characteristics of a good problem statement:• It is in interrogative form.

• It specifies relations between two or more constructs.

• It implies testability.


• Step 2: Hypotheses Formulation– Hypotheses are answers to problem statements.– Qualitative research can be used to provide

hypotheses.– Characteristics of a good hypothesis:

• It is in a conjecture form.

• It specifies relations between two or more constructs.

• It implies testability.


• Step 3: Research Design– Research design is a research plan that will

guide the research process.– Three basic research designs for social sciences

are sample surveys, experiments, and field studies.

– Three desirable characteristics of research design are precision/control, generalizability, and existential realism.


– There is no perfect research design; different research designs have different strengths and weaknesses.

• Sample surveys are high in generalizability but low in precision/control and realism.

• Experiments are high in precision/control but low in generalizability and realism.

• Field studies are high in realism but low in precision and control and generalizability.

Sample Surveys Experiments Field Studies

Generalizability High Low Low

Precision & Control Low High Low

Existential Realism Low Low High


– From hypotheses, research design can be derived:• If data are collected by the communication method

and if both the dependent and independent constructs are measurable, the research design will be a sample survey.

• If the data are collected by communication method and if the dependent constructs are measurable while the independent variables are manipulable, the research design will be an experiment.

• If the data are collected by observation in the field, the research design will be a field study.


• Step 4: Sample and Sampling– Two types of sampling: probability and non-

probability sampling.– Types of probability sampling:

• Simple random sampling

• Stratified sampling

• Cluster sampling

• Systematic sampling

• Area sampling


– Types of non-probability sampling:• Convenient sampling

• Judgmental sampling

• Quota sampling

• Snowball sampling

• Step 5: Data Collection– Two types of data collection method:

• Communication method

• Observation


– Design of measurement instruments (e.g., questionnaires)

• Uni-dimensional vs multi-dimensional constructs

• Single-item vs multiple-item constructs

• Relationships between constructs, measurement scales, measurement items, and observed variables

– Questionnaire formats: open-ended vs closed-ended questions

– Types of closed-ended questions:• Dichotomous questions

• Multiple-choice questions


• Scaled-response questions (e.g., Likert scale, semantic differential scale)

– An important issue in designing measurement instrument is the measurement property of the instrument, which includes reliability and validity.

• Reliability is the accuracy or precision of a measurement instrument: the extent that the instrument is free of error variance.

• Reliability is the internal consistency of a test.


• Validity addresses the question “Are we measuring what we want we want to measure?”

• Validity is the extent that the measurement instrument free of systematic variance and error variance.

• Step 6: Data Analysis– Statistical analyses for experiments

• ANOVA/MANOVA

• ANCOVA/MANCOVA

– Statistical analyses for surveys/field studies: dependence vs interdependence techniques


– Dependence techniques:• Simple/multiple regression

• Canonical correlation

• Structural equation modeling

• Discriminant analysis

• Conjoint analysis

– Interdependence techniques:• Correlation analysis

• Factor analysis

• Cluster analysis

• Multi-dimensional scaling (MDS)

• Correspondence Analysis


• Step 7: Manuscript Writing– Title page

– The Abstract

– The Text• Introduction

• Literature Review

• Methodology

• Results and Discussion

• Conclusion: Managerial Implication, Limitations, and Future Research Direction

– References

– Appendices

Conceptualization

Research MethodologyConceptualization

• What is Conceptualization?– See Granzin’s (1985) article on

Conceptualization of Research Projects– For a research project to be successful, it must

be properly conceptualized before the empirical portion of the research can be designed, and the data collected and analyzed.

– Conceptualization is the stage in the research project where researchers must think at an abstract level. Researcher must work with concepts and the relationships among them.


– Conceptualization is the process by which researchers, through vigorous literature review, create a conceptual model to explain a phenomenon.

– As such, conceptual model or conceptual scheme consists of a logically related set of propositions, which are stated relationships among variables.

– The most difficult part of conceptualization is how to find a good topic.


• Sources of Research Ideas– Literature review in the area of interest

• What is/are the main concept/concepts?• What are the antecedents and consequences of the

main concept? Can there be any mediator between the antecedents, the main concept, and the consequences?

• Is there any conflicting result between studies on the main concept? Can there be any moderator affecting the relationships?

• Is there any chance for further research extension?• Example: Zaichkowsky’s (1986) article on

involvement




– Literature review of basic social sciences• What is the main concept from basic social sciences

literature?

• Is the concept applicable to your field of study?

• Example: Paharia and Deshpande’s (2009) working paper on consumers buying products made from sweatshop labor based on the concept of “moral disengagement” by Bundura (1991, 1999).

– Observation of existing phenomena• What is the nature of the phenomenon observed? Is

it an antecedent of something? Is it a consequence of something?


• Is there any existing literature/studies on the phenomenon in your field? If yes, review literature thoroughly to see if you can extend existing knowledge by adding mediator, moderator, etc. If no, review literature in basic social sciences to find a proper conceptual framework for studying the phenomenon.

• Example: Oberholzer-Gee’s (2006) wrote a paper on the market for time when he observed waiting lines at the airport.

– Calls for Papers• Special issues of journals


• Sample Readings– Building conceptual models

• Zeithaml, Valarie A. (1988), “Consumer Perceptions of Price, Quality, and Value: A Means-End Model and Synthesis of Evidence,” Journal of Marketing, 52 (July), 2-22.

• Moorthy, K. Sridhar (1993), “Theoretical Modeling in Marketing”, Journal of Marketing, 57 (April), 92-106.

– Solving conflicting results• Hui, Michael K., Mrugank V. Thakor, and Ravi Gill (1998),

“The Effect of Delay Type and Service Stage on Consumers’ Reactions to Waiting,” Journal of Consumer Research, 24 (March), 469-479.


• Hui, Michael K., Xiande Zhao, Xiucheng Fan, and Kevin Au (2004), “When Does the Service Process Matter? A Test of Two Competing Theories,” Journal of Consumer Research, 31 (September), 465-475.

– Testing mediators and moderators• Hui, Michael K. and David K. Tse (1996), “What to Tell

Consumers in Waits of Different Lengths: An Integrative Model of Service Evaluation,” Journal of Marketing, 60 (April), 81-90.

– Reexamination and extension• Powpaka, Samart (2008), “Empowering Chinese Service

Employees: Reexamination and Extension”, Journal of Global Marketing, 21 (4), 271-291.


– Ideas from basic social sciences• Paharia, Neeru and Rohit Deshpande (2009), “Sweatshop

Labor is Wrong Unless the Jeans are Cute: Motivated Moral Disengagement”, Working Paper #09-079, Harvard Business School.

– Ideas from observation• Oberholzer-Gee, Felix (2006), “A Market for Time: Fairness

and Efficiency in Waiting Lines”, KYKLOS, 59 (3), 427-440.


• Criteria for Good Research Ideas– Correspondence with reality

• Good research ideas highly agree with what is accepted as true.

• Beware of “phantoms” (or invalid accepted truths)

– Coherence and parsimony• Coherence refers to whether the idea “sticks

together”.• Parsimony means a research idea should not be any

more complicated than necessary.

– Falsifiabilility• Good research ideas, if incorrect, can be falsified

by a finite set of observations.

Measurement

Research MethodologyMeasurement

• Measurement in Business Research– The Measurement Process

• In business research, the measurement process involves using numbers to represent the business phenomena under investigation.

• The empirical system includes business phenomena while the abstract system includes the numbers used to represent the business phenomena.

• Measurement process is concerned with developing a correspondence between the empirical system (e.g., preference)and the abstract system (e.g., numbers).


– Definition of Measurement• Measurement is defined as the assignment of

numbers to characteristics of objects or events according to rules.

• The nature of the relationships existing in the empirical system determines the type of numerical manipulations permissible.

– Number System Characteristics• four characteristics (of ratio scales)

– Types of Scales• There are four types of scales: nominal, ordinal,

interval, and ration scales.


• Difficulty of Measurement– Why is measurement so difficult in business

research?• Measurement is so difficult in marketing because

the phenomena of interest are typically behavioral in nature. As such, current measuring devices (e.g., questionnaires) are subject to substantial measurement error.

• Other factors of interest are concepts or constructs that are not observable. So researchers must first operationally defined the constructs and then devise a means by which they can be measured.


• Concepts of Validity and Reliability– Measurement Error

• Measurement error is minimized when a direct correspondence exists between the number system and the marketing phenomena being measured.

• In this case (idealized situations), the numbers accurately represent the characteristics being measured and nothing else.

• Most measurements possess some degree of error in that the numerical scale does not exactly represent the marketing phenomenon under investigation.

• The total error of measurement consists of two components: systematic error and random error.


– Validity and Reliability Defined• Validity of a measure refers to the extent to which

the measurement process is free from both systematic and random error.

• Reliability of a measure refers to the extent to which the measurement process is free from random errors.

• Validity is concerned with the question: Are we measuring what we think we are measuring. Reliability, on the other hand, is concerned with the consistency, accuracy, and predictability of the research findings.

• For a measure to be valid, it must be reliable.


• Types of Reliability– Test-retest reliability

• Test-retest reliability refers to the ability of the same instrument to produce consistent results when used a second time under conditions as nearly the same as possible.

• Method: Use the same instrument a second time under nearly the same conditions as possible.

– Equivalent-form reliability or alternative-form reliability

• Equivalent form reliability refers to the ability to produce similar results using two instruments as similar as possible to measure the same object.


• Method: Use two instruments that are as similar as possible to measure the same object during the same period of time.

– Internal-consistency reliability• Internal consistency reliability assesses the ability to

produce the similar results using different samples to measure a phenomenon during the same period.

– Split-half reliability• Split-half reliability is a method of assessing the

reliability of a scale by dividing into two the total set of measurement items and correlating the results.


– Cronbach Alpha• This technique computes the mean reliability

coefficient estimates for all possible ways of splitting a set of items in half.

• A lack of correlation of an item with other items in the scale is evidence that the item does not belong in the scale and should be omitted.

• Types of Validity– Face validity

• Face validity is concerned with the degree to which a measurement seems to measure what it is supposed to measure.


• Researchers judge the degree to which a measurement instrument seems to measure what is supposed to.

– Content validity• Content validity refers to the degree to which a

measurement instrument represent the universe of the concept under study.

• Content validity is the representativeness or sampling adequacy of the content of the measurement instrument.

• Content validity is ultimately a judgmental matter.


– Criterion-related validity• Criterion -related validity refers to the degree to

which a measurement instrument can predict a variable that is designated a criterion.

• Two subcategories of criterion-related validity are predictive validity and concurrent validity.

• Predictive validity is the extent to which a future level of a criterion variable can be predicted by a current measurement (predictor variable) on a scale.

• Concurrent validity is the extent to which a criterion variable measured at the same point in time as the variable of interest can be predicted by the measurement instrument (predictor variable).


– Construct validity• Construct validity is the degree to which a

measurement instrument represents and logically connects, via the underlying theory, the observed phenomenon to the construct.

• Two statistical approaches for assessing construct validity are convergent and discriminant validity.

• Convergent validity is the degree of association among different measurement instruments that purport to measure the same concept.

• Discriminant validity is the lack of, or low correlation among, constructs that are supposed to be different.


• Relevant Literature in Measurement– Churchill, Gilbert A., Jr. (1979), “A Paradigm for Developing

Better Measures of Marketing Constructs,” Journal of Marketing Research, 16 (February), 64-73.

– Gerbing, David W. and James C. Anderson (1988), “An Updated Paradigm for Scale Development and Incorporating Unidimensionality and Its Assessment,” Journal of Marketing Research, 25 (May), 186-192.

– Fornell, Claes and David F. Larcker (1981), “Evaluating Structural Equation Models with Unobserved Variables and Measurement Error,” Journal of Marketing Research, 18 (February), 39-50.

Scale Development and Validation

Research MethodologyScale Development & Validation

• Steps in Scale Development and Validation– Churchill’s (1979)

paradigm

Research MethodologyScale Development & Validation• Steps in Scale Development & Validation

(Churchill 1979)– Step 1: Specify domain of the construct

• Conduct literature review to determining exactly what is included and what is excluded in the definition of the construct.

– Step 2: Generate sample of items• Use literature review and qualitative research (e.g.,

depth interview, focus group interview) techniques to generate items that capture the domain (i.e., dimensions of the construct) as specified in Step 1.


– Step 3: Purify the measure• After the item pool in Step 2 is carefully edited,

actual data are collected to purify the items.

• The collected data are analyzed by (exploratory) factor analysis to determine the number of dimensions and to identify inappropriate items and by coefficient alpha to determine reliability and to identify inappropriate items.

– Step 4: Assess reliability with new data• Step 3 should result in face and content validity.

• After inappropriate items are deleted, data are collected for the purified items.


• Coefficient alpha or Cronbach alpha (Cronbach 1951) is the basic statistic for determining the reliability of a measure based on internal consistency.

• Other tests of reliability except test-retest reliability can also be used.

– Step 5: Assess construct validity• A new set of data is collected to establish the

construct validity (convergent and discriminant validity) and the criterion-related validity (concurrent validity and predictive validity) of the construct.

– Step 6: Developing norms• Establish the mean and SD and other statistics.

Research MethodologyScale Development & Validation• An Updated Paradigm for Scale

Development (Gerbing and Anderson 1988)– Gerbing and Anderson’s (1988) paradigm

supplements Churchill’s (1979) paradigm by adding confirmatory factor analysis (CFA) into the scale development and validation process.

– Steps in assessing the unidimensionality of the scale: exploratory factor analysis (EFA) coefficient alpha and reliability confirmatory factor analysis (CFA)

Research MethodologyScale Development & Validation• Assessing Construct Validity of a

Construct by SEM (Fornell and Larcker 1981)– Step 1: Determine if the measures have

satisfactory psychometric properties (i.e., reliability, averaged variance extracted, and discriminant validity).

– Step 2: Examine the chi square value and determine its statistical significance (i.e., the overall fit of the model).

Research MethodologyScale Development & Validation• Relevant Literature in Scale

Development & Validation– Churchill, Gilbert A., Jr. (1979), “A Paradigm for Developing

Better Measures of Marketing Constructs,” Journal of Marketing Research, 16 (February), 64-73.

– Gerbing, David W. and James C. Anderson (1988), “An Updated Paradigm for Scale Development and Incorporating Unidimensionality and Its Assessment,” Journal of Marketing Research, 25 (May), 186-192.

– Fornell, Claes and David F. Larcker (1981), “Evaluating Structural Equation Models with Unobserved Variables and Measurement Error,” Journal of Marketing Research, 18 (February), 39-50.

Research MethodologyScale Development & Validation• Sample Readings

– Richins, Marsha L. and Scott Dawson (1992), “A Consumer Values Orientation for Materialism and Its Measurement: Scale Development and Validation,” Journal of Consumer Research, 19 (December), 303-316.

– Tian, Kelly Tepper, William O Bearden, and Gary L. Hunter (2001), “Consumers’ Need for Uniqueness: Scale Development and Validation,” Journal of Consumer Research, 28 (June), 50-66.

– Bearden, William O., David M. Hardesty, and Randall L. Rose (2001), “Consumer Self-Confidence: Refinements in Conceptualization and Measurement,” Journal of Consumer Research, 28 (June), 121-134.

– Li, Hairong, Steven M. Edwards, and Joo-Hyun Lee (2002), “Measuring the Intrusiveness of Advertisements: Scale Development and Validation,” Journal of Advertising, 31 (2), 37-47.

Surveys

Research MethodologySurveys

• Definition– Survey research is the use of a questionnaire to

gather facts, opinions, and attitudes. It is the most popular way to gather primary data.

– Sample surveys are correlational studies. So survey research can establish if constructs have relationships but cannot establish cause-and-effect relationships.

• Characteristics of a Good Survey Research– A good survey is one with minimal errors.



• Types of Errors in Survey Research:– Errors in survey research include random error

(or random sampling error) and systematic errors.

– Random error - error that results from chance variation

– Systematic errors - error that results from the research design (sample design error) or execution (measurement error)

– Sample design error includes:• Frame error - error resulting from an inaccurate or

incomplete sample frame


• Population specification error - error resulting from an incorrect definition of the universe, or population, from which the sample is chosen

• Selection error - error that results from following incomplete or improper sampling procedures or not following proper ones

– Measurement error includes:• Measurement error - error that results from a

variation between the information being sought and that actually obtained by the measurement process

• Surrogate information error - error that results from a discrepancy between the information needed to solve a problem and that sought by the researcher


• Interviewer error - error that results from conscious or unconscious bias in the interviewer’s interaction with the respondent

• Measurement instrument bias - error that results from the design of the questionnaire or measurement instrument

• Processing error - error that results from incorrect transfer of information from the document to the computer

• Response errors consist of nonresponse bias and response bias

– Nonresponse bias - error that results from a systematic difference between those who do and do not respond to the measurement instrument


– Response bias - error that results from the tendency of people to answer a question falsely, through deliberate misrepresentation or unconscious falsification

• Types of Surveys– Face-to-face– Telephone interview– Direct computer interview– Self-administered questionnaires– Mail surveys– Online surveys


• Important Issues in Survey Research– Data Analysis

• Structural equation modeling tends to be used to analyze the data of survey research when the constructs are measured by multiple items.

• In this case, pay attention to the reliability and the construct validity (convergent and discriminant validity) of the constructs used in the model.

– Reliability—Cronbach alpha is higher than 0.7 (Nunnally 1978)

– Convergent validity—the proportion-of-variance-extracted index (POVEI) for each construct must be at least 0.5

– Discriminant validity—the square of the correlation between a pair of construct is lower than the POVEI of both of the constructs in the pair


• Example: How Market Orientation Affect Female Service Employees in Thailand (Powpaka 2006)




• Sample Readings– Hartline, Michael D. and O. C. Ferrell (1996), “The Management

of Customer-Contacted Service Employees: An Empirical Investigation,” Journal of Marketing, 60 (October), 52-70.

– Klein, Jill Gabrielle, Richard Ettenson, and Marlene D. Morris (1998), “The Animosity Model of Foreign Product Purchase: An Empirical Test in the People’s Republic of China,” Journal of Marketing, 62 (January), 89-100.

– MacKenzie, Scott B., Richard J. Lutz, and George E. Belch (1986), “The Role of Attitude Toward the Ad as a Mediator of Advertising Effectiveness: A Test of Competing Explanations,” Journal of Marketing Research, 23 (May), 130-143.

Experiments

Research MethodologyExperiments

• What Is an Experiment?– An experiment is a research approach in which

one or more variable is manipulated and the effect on another variable(s) observed.

– Experiments are partitioned studies.– Experimental research is often referred to as

causal research.• It is called causal research because it is the only type

of research that has the potential to demonstrate causation or cause-and-effect relationship between two or more variables.


• To demonstrate causation, that A likely caused B, we must be able to show three things: (1) concomitant variation, (2) appropriate time order of occurrence, and (3) elimination of other possible causal factors.

• Experimental Validity– Internal and external validity

• Validity of a measure refers to the degree to which the measure is free from both systematic and random error.

• In addition to the general concept of validity, in experimentation, there are two specific kinds of validity: internal validity and external validity.


• Internal validity is the extent to which competing explanations for the experimental results observed can be avoided.

• External validity is the extent to which causal relationships measured in an experiment can be generalized to outside persons, settings, and times.

– Extraneous variables: threats to experimental validity

• History refers to any variable or event other than those manipulated that takes place between the beginning and end of the experiment and that might affect the value of the dependent variable.


• Maturation refers to changes in subjects that take place during the experiment that are not related to the experiment but may affect their response to the experimental factor.

• Instrument variation refers to any differences or changes in measurement instruments (e.g., interviewers or observers) that explain differences in measurements.

• Selection bias refers to systematic differences between the experimental group and control group because of a biased selection process.

• Mortality refers to the loss of test units or subjects during the course of an experiment.


• Testing effect is an effect that is a by-product of the research process and not the experimental variable.

• Regression to the mean refers to the tendency for behavior of subjects to move toward the average for that behavior during the course of an experiment.

• Experimentation: Summary of Basic Concepts– Experimental design and treatment

• Experimental design is a test in which the researcher has control over one or more independent variables and manipulates them.

• Treatment is the independent variable that is manipulated.


• Manipulation refers to the process in which the researcher sets the levels of the independent variable to test a particular causal relationship.

– Experimental effects• Experimental effect is the effect of the treatment

variable on the dependent variable.

– Control of other (extraneous) causal factors• Extraneous causal factors are also referred to as

confounding variables.

• Four basic approaches are used to control extraneous factors: (1) randomization, (2) physical control (e.g., matching, mode K), (3) design control, and (4) statistical control (e.g., ANCOVA).


• Types of Experimental Design– Three pre-experimental designs

• One-shot case study

• One-group pretest-posttest design

• Static-group comparison

– Three true experimental designs• Posttest-only control group design

• Pretest-posttest control group design

• Solomon Four-Group design


– Quasi-experimental designs• Interrupted time-series design

• Multiple time-series design

• Nonequivalent control group design

• See more designs in Cook and Campbell’s (1979) Quasi-Experimentations

• The Experimental Setting—Laboratory or Field Experiments?

• Laboratory experiments are experiments conducted in a controlled setting.

• Field experiments are tests conducted in an actual market environment.


• Issues in Experiments– Manipulation of the independent variables

• Realistic

– Manipulation checks• Need manipulation-check variables

• Successful manipulation means:– The means of the manipulation-check variable of an

independent variable under different conditions are significantly different as required

– No confounding effect

– The effects of the interaction between the independent variables on manipulation-check variables are not significant


– Measurement• The dependent variables are measureable (or

metric).

• Pay attention to the construct validity (convergent and discriminant validity) of the dependent variables.

– Data Analysis• ANOVA/MANCOVA

• ANCOVA/MANCOVA

• ANOVA/MANCOVA/ANCOVA/MANCOVA by structural equation model (Bagozzi and Yi 1989)


– Results and Interpretation• Main Effects—the effect of each of the independent

variable on the dependent variable(s)

• Interaction Effects—the effect of an independent variable on the dependent varible(s) depends on the level of another dependent variable

• Types of interaction effect– Disordinal interaction

– Ordinal interaction


• Example: Hui et al’s (2004) Study 3 on Empowerment Across Culture



• Manipulation: Service scenarios


• Manipulation checks: Pilot studies– Manipulation-check variables

• Power distance for nation (Pilot Study 1)

• Perceived discretionary power for empowerment (Pilot Study 2)

• Perceived good reason for request for request nature (Pilot Study 2)

– Note• The authors only test if the means of the

manipulation-check variable are significantly different between different conditions as required. There are no confounding and interaction tests.


• Sample Readings– Hui, Michael K. and David K. Tse (1996), “What to Tell

Consumers in Waits of Different Lengths: An Integrative Model of Service Evaluation,” Journal of Marketing, 60 (April), 81-90. (Read only Introduction and Literature Review)

– Hui, Michael K., Mrugank V. Thakor, and Ravi Gill (1998), “The Effect of Delay Type and Service Stage on Consumers’ Reactions to Waiting,” Journal of Consumer Research, 24 (March), 469-479. (Read only Introduction and Literature Review)

– Hui, Michael K., Xiande Zhao, Xiucheng Fan, and Kevin Au (2004), “When Does the Service Process Matter? A Test of Two Competing Theories,” Journal of Consumer Research, 31 (September), 465-475. (Read only Introduction and Literature

Review)

Multivariate Data Analysis

Research MethodologyMultivariate Data Analysis

• Types of Multivariate Techniques– Dependence Techniques

• Multiple regression analysis• Canonical correlation• Multivariate analysis of variance and

Covariance• Discriminant analysis• Conjoint analysis• Structural equation modeling


• Types of Multivariate Techniques– Interdependence Techniques

• Factor analysis• Cluster analysis• Multidimensional scaling• Correspondence analysis


• Multiple Regression– Multiple regression is appropriate when

there is a single metric dependent variable and two or more metric dependent variable.

– The objective of multiple regression is to predict the change in the dependent variable in response to changes in the independent variables.


• Canonical Correlation– Canonical correlation can be viewed as a

logical extension of multiple regression analysis.

– Canonical correlation correlates simultaneously two or more metric dependent variables and two or more metric independent variables.


• Multivariate Analysis of Variance and Covariance– MANOVA is a statistical technique for

simultaneously explore the relationship between two or more metric dependent variables and one or more categorical variables.

– MANCOVA removes the effect of “covariates” on the dependent variables.


• Multiple Discriminant Analysis– Multiple discriminant analysis is

appropriate when there is a single categorical dependent variable and two or more metric independent variables.

– The objective of MDA is to understand group differences and to predict the membership of an entity (individual or object).


• Conjoint Analysis– Conjoint analysis is appropriate when

there is a single ordinal dependent variable and two or more categorical variables.

– Conjoint analysis provides “utilities” (or relative importance) for each level of each categorical variable.


• Structural Equation Modeling– Structural equation modeling

simultaneously analyzes a set of simple and multiple regression functions.

– Structural equation modeling consists of (1) the structural model and (2) the measurement model.

– The structural model represents the relationship among the latent variables.


– The measurement model represent the relationships between the observed variables (or “indicators”) and the latent variables.

– LISREL is one of the most popular software package for structural equation modeling. Other popular software packages include AMOS (by SPSS) and EQS.


• Factor Analysis– Factor analysis, including principal

component analysis and common factor analysis, is a statistical technique to analyze interrelationships among a large number of metric variables and to explain these variables in terms of their common underlying dimensions (or factors).


• Cluster Analysis– Cluster analysis is an analytical technique

for developing meaningful subgroups of individuals or objects based on the similarities among the entities.

– Cluster analysis groups individuals or objects into groups (or “clusters) based on metric variables while factor analysis groups metric variables into factors.


• Multidimensional Scaling– Multidimensional scaling transforms

similarity/dissimilarity scores of pairs of objects into distances represented in multidimensional space.

– The resulting perceptual maps show the relative positions of all objects. The more similar a pair of objects are, the closer the two objects are.


• Correspondence Analysis– Correspondence analysis facilitates both

dimensional reduction of object ratings on a set of attributes and the perceptual mapping of objects relative to these attributes.

– In most basic form, correspondence analysis employs a contingency table (cross-tabulation of two categorical variables)

Structural Equation Modelingby LISREL

SEM by LISREL• What is Structural Equation Modeling?

– SEM is a multivariate technique combining aspects of factor analysis and multiple regression that allows an investigation of the structure of relationships among the observed variables and latent variables and those among the latent variables.

– SEM models consist of the measurement model (representing the relationships between each latent variable and its observed variables) and the structural model (representing the relationships among the latent variables).

SEM by LISREL– The measurement model of an SEM model

must be acceptable before the result of the structural model can be interpreted.

– In a good measurement model, each construct must have (1) high reliability (Cronbach α > 0.7), (2) construct validity (convergent validity and discriminant validity), and (3) acceptable overall model fit (e.g., relative fit indices > 0.90).

– Widely used SEM programs include LISREL, EQS, and AMOS.

SEM by LISREL• The Objective of Using SEM

– Confirmatory Factor Analysis (or CFA)– Strictly Confirmatory (or SC) Situation– Alternative Model (or AM) Situation– Model Generation (or MG) Situation– Testing Equality of Coefficients– Path Analysis

• Type of SEM Analysis– Single-Group SEM– Multiple-Group SEM– MANOVA by SEM

SEM by LISREL• Steps in Using LISREL

– Step 1: Save the SPSS data set (.sav) into the tab-delimited data set (.dat)

– Step 2: Write a PRELIS 2 command file (.pr2) to analyze the tab-delimited data set (.dat) in order to obtain the appropriate moment matrix (.cov) for further analysis by LISREL 8.

– Step 3: Write a LISREL 8 command file (.ls8) to analyze the matrix (.cov) in order to obtain the results (i.e., the measurement model and structural model).

SEM by LISREL• Saving Data File from SPSS Data Set

– Step 1: In the SPSS data file, click “File” and then click “Save As”. The “Save Data As” window will be opened.

– Step 2: Change “Save as type” from “SPSS (*.sav)” to “Tab-delimited (*.dat)”.

– Step 3: Type your file name into the box of “File name”.

– Step 4: Uncheck the box of “Write variable names to spread sheet”.

– Step 5: Click “Save”.

SEM by LISREL• How to Write PRELIS Command File

– Title line: Title of the research project– Data line: DA NI=n MI=m (n is the number of

observed variables and m is the number representing the missing value)

– Label line: LA (followed by the list of observed variables in the next line)

– Raw data line: RA=location of data (.dat) file– Continuous variables line: CO (followed by list

of variables or ALL)

SEM by LISREL– Selective delete line (option line): SD (followed

by the list of variables to be deleted from the analysis)

– Select case line (option line): SC (followed by the variable and the value selected for the analysis)

– Output line: OU MA=CM SM=location for saving the covariance matrix (.cov)

SEM by LISREL• How to Write LISREL Command Files

– Title line: Title of the research project– Observed variables line: Observed variables:

(followed by the list of observed variables)– Covariance matrix from file line: Covariance

matrix from file (followed by the location of the .cov file from PRELIS)

– Sample size line: Sample size n (n is the size of the sample or subjects)

SEM by LISREL– Latent variables line: Latent variables:

(followed by the list of latent variables)– Relationships line: Relationships: (followed by

the measurement model and then the structural model in the next lines)

– Path diagram line: Path diagram– Admissibility check line: Admissibility=off– End of problem line: End of problem

Conclusion

Research MethodologyConclusion

• Advice from Former Editors of Journal of Marketing– Varadarajan, P. Rajan (1996), “From Editor: Reflections

on Research and Publishing,” Journal of Marketing, 60 (October), 3-6.

– Stewart, David W. (2002), “Getting Published: Reflections for an Old Editor,” Journal of Marketing, 66 (October), 1-6.