A Comparison of Modeling Strategies for Latent Class Analysis With Mixed-Metric Indicators

• In Event: 61.084 - Criteria and Bias in Measuring Latent Constructs and Classes

Abstract

This simulation study examines the performance of fit indices commonly used by applied researchers interested in latent class analysis (LCA) using mixed-metric data. Despite the advantages of LCA over traditional clustering, model selection aided by fit indices remains a significant challenge to researchers. Conditions will be generated to mirror conditions found in applied educational and psychological research. The accuracy with which common fit indices identify the true three-class LCA model is examined while varying indicator type (i.e., continuous and categorical), sample sizes, class prevalence, and model type. This study will examine fit index performance between two model types: 1) LCA models with categorical or continuous indicators and covariates predicting class membership, and 2) LCA with simultaneous inclusion of mixed-metric indicators.

Authors

• Kevin Wells, Baylor University

• Kari Hodge, NACE International Institute

• Grant B. Morgan, Baylor University