ABSTRACT
Structural equation models (SEM) have become an increasingly popular technique for analysis of developmental research questions. However, a number of unfortunate misconceptions can be found in the literature regarding the nature, potential, and pitfalls of SEM. It is fallacious to assume that use of SEM techniques guarantees sound causal inference from correlational data; it is equally fallacious to argue that use of SEM for purposes other than testing causal models is an invalid misapplication of the method. In developmental research, important descriptive research questions can be shown to be linked to SEM models in two important ways: Alternative SEM models may be used to provide direct statistical tests of important descriptive developmental hypotheses, and SEM model parameters can be interpreted with respect to fundamental issues in developmental analysis (e.g., estimating the degree to which differential developmental patterns alter distributions of individual differences). This paper develops the logic and procedures for implementing longitudinal SEM techniques to address descriptive developmental questions, with a brief illustration of the application of SEM to longitudinal factor analysis.
