Scaling Models for Categorical Variables: An Application of Latent Structure Models

Abstract

We demonstrate how latent structure analysis can be used to model rating scale data that have ordered categories, and propose a modification of the basic latent structure approach to analyze response errors in the context of a traditional multitrait-multimethod matrix. The extended approach provides the researcher with the ability to (1) use formal test statistics to select a response error model, (2) assess the effects due to traits versus methods, and (3) examine and test a wide array of plausible measurement error hypotheses.