A Bayesian Approach to Multilevel Latent Growth Model (Poster 12)

• In Event: AERA Poster Session 4
  In Poster Session: Poster Session (San Diego)

Abstract

Multilevel Latent Growth Model (MLGM) is used to estimate the growth trajectories of individuals nested within clusters or groups. It is often estimated from a frequentist framework using Maximum Likelihood (ML), which often results in inaccurate parameter estimates and convergence problems due to the complexity of the multilevel model. The Bayesian approach of using Markov chain Monte Carlo estimation method can be incorporated to improve the accuracy of parameter estimation and overcome convergence issues. This simulation study compared the performance of the Bayesian and frequentist approaches in estimating MLGM with covariates. Results revealed that the Bayesian estimation provided more accurate parameter estimates, especially under small sample size, and reduced estimation problems of non-convergence or negative variances.

Authors

• Suyoung Kim, University of Texas at Austin

• Jiwon Kim, University of Texas at Austin