Detecting differential effects through a moderation model

• In Event: AERA Roundtable Session Sunday 11:45 am - Gold 4
  In Roundtable Session: Advanced SEM Approaches for Modeling Interaction, Heterogeneity, and Variability

Abstract

While regression-based and machine learning approaches have been used to examine heterogeneous moderation effects, regression mixture models (RMMs) remain underexplored in this context. This simulation study is the first to evaluate whether RMMs can reliably detect and estimate such differential effects. Using Monte Carlo methods, we varied sample size, mixing proportion, predictor correlation, and effect sizes across 48 conditions, each with 500 replications. Preliminary findings from three conditions suggest that moderation parameters can be estimated with minimal bias under certain conditions. However, model fit indices did not consistently identify the correct number of latent classes. These results highlight both the promise and limitations of RMMs for modeling subgroup-specific moderation effects.

Authors

• Min Liu, Baylor University

• Grant B. Morgan, Baylor University

• Gregory R. Hancock, University of Maryland

• Chih-Pu Dai, Texas A&M University