Selecting the Correct RI-CLPM Using Chi-Square-Type Tests and AIC-Type Criteria

Abstract

In the field of behavioral and social science, the random intercept cross-lagged panel model (RI-CLPM) is increasingly gaining popularity among researchers. However, a challenge is the selection of the appropriate RI-CLPM type, more specifically, the selection of the random intercepts (RIs) in the model. This study aims to address this concern by comparing four techniques: the Chi-square difference test, the Chi-bar-square difference test, Akaike’s information criterion (AIC), and the recently developed AIC-based criterion called Generalized Order-Restricted Information Criterion Approximation (GORICA). The results demonstrate the effectiveness of each technique in selecting the correct model under various simulation conditions. In the case of selecting a true RI-CLPM with one random intercept, the Chi-bar-square difference test demonstrates the highest performance. For other scenarios, the GORICA surpasses other techniques.