The performance of latent growth curve model-based structural equation model trees to uncover population heterogeneity in growth trajectories

Abstract

Behavioral researchers have shown growing interest in structural equation model trees (SEM Trees), a new recursive partitioning-based technique for detecting population heterogeneity. In the present research, we conducted a large-scale simulation to investigate the performance of latent growth curve model (LGCM)-based SEM Trees for uncovering between-individual differences in patterns of within-individual change. Simulation results showed that the correct estimation rates of the number of classes are most strongly related to the agreement rate of the covariate with its true latent profile, and the number of true classes also has a serious negative impact on correct estimation rates of the number of classes. SEM Trees is not always sensitive to the influence of model misspecification, and its impact differs according to a complex function of the types of misspecification as well as the statistical properties of the template model. On the whole, LGCM-based SEM Trees is a robust and stable approach under possible model misspecifications.