ABSTRACT
Structural equation modeling (SEM) is a method of multivariate analysis concerned with representing population moments on the basis of relatively few parameters that are hypothesized on the basis of a substantive theory. Historically, the main emphasis in SEM has been to model centered second-order moments, namely covariances, and the associated field is often called covariance structure analysis. SEM is used in many disciplines and its popularity is due to its ability to model complex behaviors of variables according to a theory, especially in contexts where it is hard or impossible to use randomized experiments to evaluate causal theories. The broad attraction of SEM is that it incorporates methodology from psycho-metrics, econometrics, and sociometrics. Irregardless of model structure chosen to be used, in order for any latent variable to be identified in SEM, the model structure must imply a unique parameterization.
