Information Complexity and Several Other Information Theoretic Criteria in the Selection of Structural Equation Models

• In Event: Measurement Models

Abstract

This study applies Information Complexity (ICOMP) criteria to model evaluation and selection in structural equation models (SEM). To that end, the study follows an existing simulation protocol from the SEM literature and examines the performance of ICOMP in identifying the true simulation model out of a few competing models under a variety of experimental conditions. The study takes advantage of the lavaan package in R to code ICOMP and a few other fit statistics (information theoretic measures, in particular) commonly used in SEM. The study compares the performance of ICOMP with that of other selection criteria through large scale Monte Carlo simulations. The study serves as the foundation for the implementation of ICOMP in more general structural equation models.

Authors

• Hamparsum Bozdogan, The University of Tennessee

• Kenneth A. Bollen, University of North Carolina

• Hongwei Yang, University of Kentucky