ABSTRACT
This book contains several excellent chapters on structural equation models for longitudinal data. For the most part, the approaches described in those chapters and elsewhere are based on an implicit assumption that any latent variables are continuous. By contrast, this chapter discusses a different point of view on latent variables. In the point of view described here, latent variables are discrete. Individuals are not arranged along a latent continuum, as they are in continuous latent variable models, but rather placed into latent categories. Just as in factor analysis, the latent variable must be measured by means of manifest or directly observed variables, often referred to as indicators. However, here the indicators are discrete categorical variables. This measurement model is known as latent class theory (e.g., Goodman, 1974).
