Theory and Practice of Structural Equation Modeling

Prof. Albert Satorra
Universitat Pompeu Fabra
Departament d'Economia i Empresa
Barcelona
2.-10.12.2010

Course description

Structural Equation Modeling (SEM) is a statistical modeling technique to assess hypothesis of relationships among variables. A key feature of SEM is that unobserved variables (latent constructs) are contemplated in the model. Latent variables correspond to concepts that have content based on theory. Relationship among variables (latent and/or observable) are postulated into the form of a model. Model fit and goodness of fit testing exposes substantive theory to empirical disconfirmation.

The course focus on the theory and applications of SEM to social and behavioural sciences, bringing up examples in marketing, management and business studies. The course cover topics such as confirmatory factor analysis, measurement and structural models, path analysis, mediation, latent growth modeling, assessment of model fit and, finally, the implementation SEM in the particular software EQS.

• Required Course Reading: Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd Ed). New York: Guilford Press.
• Required Software: EQS and R

Philosophy of the course:

The course is structured like a workshop. I will familiarize you with some applications of structural equation modeling and give you some direct experiencing using the software package EQS to estimate the models. The free software R will also be used, and I will be most happy to help you to move to using other software if interested, such as Mplus, LISREL, EQS, or AMOS, and others. After this class, you will not be an expert on the topic; i fact, far from it! you will learn just enough in this class to make you dangerous, for there is much more to structural equation modeling than knowing how to set up a model and clicking the right button on a computer. I assume that you will use this introduction to help progress yourself in using knowledge and common sense in the practice of multivariate analysis.

Topics to be covered:

• Regression refresher, basic concepts, use of software
• Path analysis with observed variables
• Measurement, exploratory, and confirmatory factor analysis
• Structural models with latent variables
• Estimation, testing and modification of models
• Latent growth models
• Multiple group models
• SEM for multilevel data (if time available)

Evaluation:

For those taking the course for credit, grades will be based on attendance and discussion in class and the completion of assignments or other busy work. The 50 % of the final grade will be based on a data analysis project that you complete using either your own data or data available to you through an advisor or through a public archive. You should write this paper as if you were planning on submitting it to an academic conference or journal. This assignment can be worked out in a group of at most two students.

Materials:

• What is structural equation modelling (SEM) whatisSEM
• Regression as moment structures and factor analysis regresmom
• Errors in variables errorinvariables
• Simultaneous equations simultaneous
• The LISREL model SEMbasicmodels

---

Last change: 2011-04-08 by mp