Nonnormality and a Small Latent Class in Growth Mixture Modeling

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

Abstract

Nonnormality in educational data is prevalent. With nonnormal data, the extraction of a spurious class is of concern in GMM due to the violation of the normality assumption. This study demonstrates that those spurious classes tend to be very small. A critical challenge in applied educational research is distinguishing spurious small classes from true small classes that may need unique educational needs. This study investigates whether true small classes in the population can be distinguishable from spurious small classes and how well robust GMM (nonnormal GMM) can identify true small classes accurately through Monte Carlo simulations. Based on the results, practical implications and guidelines are provided for educational researchers in evaluating small classes.

Authors

• Emma Elísa Evudóttir, University of South Florida - Tampa

• Eunsook Kim, University of South Florida

• Courtney Howard-Kirby, University of South Florida

• Xilong Jing, University of South Florida

• Xin Tong, University of Virginia

• Yan Wang, University of Massachusetts Lowell

• David Goretzko, Utrecht University