Structural Equation Models

Abstract

Listen

Translate article

The purpose of this chapter is to present an accessible overview of recent research on what are termed structural equation models (SEM). This presentation is intended for graduate level students in the behavioral sciences, possibly taking a SEM class, but formal algebra or calculus is not required. First, SEM is broadly defined, and the increasing use of this approach to data analysis is described. In general, SEM techniques are increasingly used in the behavioral and social sciences. Second, some technical features of SEM are presented to illustrate key benefits of SEM. Some classical issues are described that highlight issues SEM researchers usually find to be important, and the big appeal of SEM comes when if offers some hope to deal with these issues. Third, we consider the inclusion of common factors as latent variables in path models that can be incorporated into SEM. We claim that the inclusion of common factors is what really makes SEM different than other statistical approaches. Fourth, we describe how SEM calculation works, and this gives rise to various indices of goodness-of-fit. Many researchers herald these techniques, although this seems to be a leftover from prior statistical training. Fifth, we provide an illustration of contemporary data generation and computer programming (using CALIS, Mplus, and OpenMx). In the final section, we illustrate some options from our previous SEM work, answer specific questions about SEM practices, and include a discussion of issues for future SEM uses.