Multilevel Growth Mixture Models for Classifying Groups

This article introduces a multilevel growth mixture model (MGMM) for classifying both the individuals and the groups they are nested in. Nine variations of the general model are described that differ in terms of categorical and continuous latent variable specification within and between groups. An application in the context of school effectiveness research is presented. Schools are classified into three Type B effectiveness categories based on their mean student mathematics achievement growth trajectories, controlling for differences in students' backgrounds across schools. The classification outcome is regressed on a set of school practice variables to investigate the association between practices and cognitive development. Various issues related to model specification are discussed, including the use of covariates to identify substantively meaningful classes.