Piecewise Growth Mixture Models for Classifying Growth Trajectories: Maximum Likelihood/Expectation Maximization Versus Bayesian Estimation

• In Event: 65.060 - Multilevel Modeling: Nonnested Multilevel Data, Measurement Invariance, and Bayesian Approaches

Abstract

Piecewise growth mixture modeling (PGMM) can be used to analyze longitudinal data that are hypothesized to have heterogeneous growth trajectories and exhibit the discontinuity of growth trajectories of distinctive turning points at which growth rate changes. The flexibility and dynamism of the method has attracted enthusiastic applications in social and behavioral science. However, up to date, there has been no simulation study to examine performance characteristics of PGMM, particularly, in the accuracy of identifying the number of latent growth trajectories. This simulation study aims to provide reasonable ways for how the model fit statistics examined could be used in combination in specific conditions to decide an optimal PGMM under the two prevailing paradigms: the Maximum likelihood estimation and Bayesian estimation.

Authors

• Ling Ning, University of California - Davis

• Wen Luo, Texas A&M University - College Station