Effects of Mixing Metrics and Distributions Simultaneously in Structural Equation Modeling: A Simulation Study

• In Event: Advanced Topics in Structural Equation Modeling

Abstract

Mixing continuous and categorical variables in a model is not uncommon in applied research, particularly when testing full structural models; however, no comprehensive simulation studies have systematically investigated the effects of nonnormality using both continuous and categorical observed variables simultaneously in an SEM model. This study compared the performance of two estimators (robust ML and robust WLS) in a small and large covariance model using a mixture of continuous and categorical variables. To emulate real data applications, mixtures of data types and data distributions between and within latent constructs were manipulated. Experimental conditions also considered various sample sizes. Performance was evaluated according to several criteria including parameter estimate bias, standard error bias, and bias of several tests of model fit.

Authors

• Cindy A. Kronauge, University of Northern Colorado

• Susan Hutchinson, University of Northern Colorado