In this paper we examine and compare the factor analytic and grade of membership (GoM) models for characterizing the latent structure of a set of indicators, where the indicators measure a smaller set of distinct dimensions. We describe the factor analytic and grade of membership models statistically and then evaluate their usefulness under various circumstances that arise in social science research. A comparison between factor analysis and GoM is of particular interest because factor analysis is now used extensively in sociology, whereas GoM is a relatively new procedure originally developed for the analysis of symptoms of mental and physical disability. After discussing the generic differences between factor analysis and GoM, we illustrate the differences with an empirical example that focuses on indicators of job values held by U.S. high school seniors in 1991.

* University of Minnesota

** American Institutes for Research