The Impact of Replacing First-Order Latent Factors with Composite Scores on Model Fit and Structural Estimates in SEM

• In Event: Beyond Traditional SEM: Model Fit, Heterogeneity, and New Paradigms

Abstract

This study aims to address this critical gap by systematically examining how replacing first-order latent factors with composite scores affects structural parameter estimates and model fit in SEM. This research aims to provide empirical evidence to inform applied researchers and methodologists about the risks and conditions under which this modeling simplification might lead to misleading results. The simulation design factors included sample size, the magnitude of factor loadings, and the magnitude of structurl coefficients. Preliminary results showed that the RMSE of estimating structural relationships was comparable between full SEM model and SEM model with composiste scores. When sample size was small, the model fit in SEM model with composite scores was better than that in the ful SEM model.

Authors

• Chunhua Cao, University of Alabama

• Justice Dadzie, University of Alabama

• Theode Niyirinda, University of Alabama