This course is designed for individuals who want to become familiar with the statistical techniques known collectively as “latent variable modeling”. Throughout the course, widely available, but specialized, software such as LISREL or MPlus will be used for the computation. The course focuses on the class of techniques and statistical theory that include, for example: (a) the class of models referred to as LISREL models or structural equation models (SEM), (b) unrestricted maximum likelihood factor analysis, (c) path or causal models, and (d) confirmatory factor analysis. The course will also include discussion, and examples, of measurement invariance, and rating scales – Likert response data. Both exploratory and confirmatory modeling strategies will be discussed, with an emphasis on the statistical and confirmatory approach.
1. Basic ideas of latent variable modeling and factor analysis.
2. Overview of “exploratory” factor models
3. General Linear Latent Variable Model
4. Confirmatory Factor Analysis (CFA)
5. Applications of CFA:
a) Construct validity and measurement studies
b) Multi-group CFA
c) Test theory models of “equivalence”, sets of congeneric tests
d) Matters of factorial and measurement invariance
e) Models for latent growth (optional)
6. Special Problems
a) Dealing with binary and rating scale (or Likert) data and how factor analysis is related to Item Response Theory (IRT)
b) Incomplete data (i.e., missingness)
c) Cautions regarding “causal” modeling; what is “causal” about causal modeling?
d) The matter of equivalent models
e) Model identification
f) Methods for setting the metric of the latent variable(s)
Prerequisites: Successful completion of EPSE 592 and EPSE 596, or at least two courses (one of which must be a graduate course) in statistics and/or data analysis. It will be very helpful to have taken (or be taking) a course in regression because some of those concepts will be built on in this class. If you are unsure whether your background is sufficient, please contact the instructor.