S651 | 3713 | Scott Long

Structural equation modeling (SEM) is a general method for analyzing the relationship among variables. In its most complex form, SEM allows you to have multiple equations with errors in equations and errors of measurement for all the variables in the model. The course will approach this very general model by progressing through a series of special cases that are quite useful and powerful in and of themselves. First, we will consider simultaneous equations systems for multiple regression. In this form of the model, all of the variables are measured without error. Next, we consider the case in which all variables are measured with error, but that there is a structural relationship among the variables. This form of SEM is referred to as confirmatory factor analysis. The advantage of confirmatory factors analysis compared to exploratory factor analysis is considered. Finally, we will consider the full model in which simultaneous equation systems and measurement error are combined. Note that the SEM goes by various names, including covariance structure analysis and LISREL modeling. This is a graduate level class in applied statistics that assumes familiarity with the linear regression model as taught in Sociology 554. Students will be required to write a substantive paper using structural equation modeling, or a methodological paper dealing with a problem in SEM.