MKT. 6355

THEORY TESTING

Description:

This course will examine the interface between the philosophy of science underlying theory testing and statistical methodologies for testing theories. We will explore measurement issues (classical test theory, exploratory and confirmatory factor analysis, and item response theory), statistical inference and hypothesis testing, path analysis and structural equations modeling from both an econometric perspective (that is, issues of exogeneity and instrumental variables) and the latent variable structural equations modeling perspective (LISREL-type models). Special attention will be given to the assessment of causality. Issues regarding experimental and quasi-experimental designs and analysis will be included. The course is designed to provide the student with a sound basis for conducting original research and with the background for studying more specialized and advanced techniques.

The course will consist of both lecture and discussion of assigned readings. A paper describing the analysis of new data or a re-analysis of existing data constitutes the primary requirement for the course. Some of the readings are available at http://rhowell.ba.ttu.edu under the heading “Readings for theory testing”. For further information, contact roy.howell@ttu.edu.

Objectives:

Upon completion of the course, the students will be able to:

– recognize situations in which structural equation modeling may be useful,

– demonstrate a general understanding of structural equation modeling techniques,

– use structural equation modeling techniques in their research

– explain the limitations of structural equation modeling techniques, and

– critique articles that use structural equation modeling techniques.

 

Project  Students will critique a published article that uses structural equation modeling. The critique will focus on the analyses used in the article. It should discuss whether the use of structural equation modeling is appropriate/inappropriate. The critique should also include a discussion of the properties and development of the measures used in the study. In addition, the student will try to reproduce the analysis undertaken in the article. (Therefore, the article needs to include either a correlation matrix or a covariance matrix.)

Alternatively, students may perform structural equation modeling on a data set of their choice. Briefly summarize the rationale for the model, and write-up the results and conclusions in journal format.

 

Abbreviated class notes available at: class1 class2

Grading:

Class participation and discussion                   40%

Project                                                                   60%

 

Tentative Class Schedule CHECK WEEKLY FOR UPDATES!!

Week 1                  Introduction

                                Basics, History

 

Week 2                  A philosophy for theory testing        Meehl article  Review article

 

                                Exploring SEM                                     Bring an article employing SEM for discussion

 

Week 3                  Path Analysis                      

                                Models imply covariances

Testing fit                                              Excel example

 

Week 4                  Adding the measurement model      Anderson and Gerbing article

                                Latent variables and CFA                 Borsboom articles – latent variables, validity

                                                                                                 Validity2

Construct lecture

 

Week 5                  LISREL programming and output   run ex56.ls8 from LISREL manual and bring output

                                Fit indices                                              Week 6                  Model modification

Mediation analysis                          alpha mediation1 mediation2

 

Week 6                  Fit revisited                           Fit1         Yuan                      Saris

                                Moderation                           Little

 

Week 7                  Normality              MLGLS   Time

                               

                                Missing Data        Missing 1               Missing 2                              

 

Week 8                  Mean Structures

                                Ordinal Data         Ordinal

                                Multiple Group Analysis    BSM                       Invariance

                                Multilevel

 

 

Week 9                  Causation              Pearl      Rubin Causal Claims  New-reading

                                Omitted variables,

                                2SLS                      Wooldrige

                                Search for ‘Heckman selection bias’ and ‘propensity score matching’ & prepare to discuss

 

Week 10                Reporting SEM analyses                    Reporting 1  Reporting 2

                                Revisit the article you discussed the second week of class. How do you feel about it now?