Unifactor Latent Trait Models Applied to Multifactor Tests: Results and Implications

Abstract

In order to extend the use of latent trait models across the full spectrum of mental testing, the applicability of the models to multivariate data must be determined. Since all of the commonly used models assume a unidimensional test, the applicability of the procedures to obviously multidimensional tests, such as achievement tests, is questionable. This paper presents the results of the application of latent trait analyses to a series of tests that vary in factorial complexity. The purpose is to determine what characteristics are estimated by the models for these tests, while at the same time determining the relationship of latent trait parameters to traditional item analysis and factor analysis indices.