Incorporating Covariates in Bayesian Piecewise Growth Mixture Models

• In Event: Methodological Considerations and Guidance for Estimating Longitudinal/Growth Models

Abstract

This study presents a methodological extension to the existing Piecewise Growth Mixture Models (PGMM) that are used in educational and psychological literature to analyze longitudinal data. The model under consideration allows the incorporation of covariates to inform latent class membership by integrating the existing model with a multinomial logistic regression. A random sample of 1000 individuals from a national survey was used to motivate the applicability of this method. The data were analyzed using the PGMM with time-invariant covariates – sex and dropout status. The empirical results show that a two-class model fits the data best with Class 1 showing a linear trend and class 2 showing piecewise trend. The covariates were shown to be statistically significant predictors for class membership.

Authors

• Rik Lamm, Bloomington Public Schools

• Nidhi Kohli

• Eric F. Lock, University of Minnesota