A Comparison of Methods for Exploring Heterogeneous Subgroups Using Covariates in Longitudinal Study

Exploring heterogeneous subgroups in longitudinal studies is a crucial task. The use of SEM Trees, based on the decision tree paradigm, has recently been expanding. However, the application of SubgroupSEM, which is based on the subgroup discovery paradigm, is still insufficient. The purpose of this study is to compare the analysis procedures and results of the two methods from a theoretical perspective and discuss their applicability. To achieve this, we conducted an analysis using mathematics vertical scale scores and eight covariates provided by the Gyeonggi Education Panel Study (GEPS). The results revealed differences in the partitioning methods of the covariate space, the covariates used to derive the subgroups, and each subgroup membership. Based on these comparison results, this study provided practical guidelines to help researchers appropriately apply each method.