A Factored Regression Approach to Modeling Interactions Between Latent and Multicategorical Variables

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

Abstract

Latent variable moderation is crucial for modeling interactions involving unobserved constructs, yet few methods handle categorical moderators. This study evaluates a new factored regression approach that estimates latent-by-categorical interactions. Using Monte Carlo data simulations, we varied sample size, indicator loading size, number of groups, and interaction effect size to compare factored regression with multigroup SEM and manifest-variable regression. Outcomes include bias, mean squared error, interval coverage, and power. Results clarify conditions under which factored regression offers advantages over existing methods. A real data example using the rblimp R package will be developed to demonstrate model implementation and interpretation. This work expands applied researchers’ options for estimating interactions involving latent variables and categorical groupings.

Authors

• Remus Rose Mitchell, University of California - Los Angeles

• Craig K. Enders, University of California - Los Angeles