A Comparison of Parameter Recovery in Multilevel Modeling Estimation: Frequentist Versus Bayesian Estimation

• In Event: 49.061 - Advanced Modeling Strategies for Longitudinal, Multilevel, and/or Survey Data

Abstract

This simulation study was designed to compare the parameter recovery between frequentist and Bayesian estimation in multilevel modeling. Specifically, frequentist estimation via iterated generalized least squares (IGLS) and Bayesian estimation via the Markov chain Monte Carlo (MCMC) were used to test the accuracy of multilevel modeling parameter estimation. Additionally, two Bayes point estimates were compared: the MCMC posterior mean and the MCMC posterior median. In smaller number of level two units, level two variance component IGLS estimates with less than 50 groups were substantially under-estimated and MCMC posterior mean with less than 30 groups were substantially over-estimated. In general, the use of MCMC posterior median resulted in less biased parameter estimates as compared to IGLS and MCMC posterior mean estimates.

Authors

• Hyewon Chung, Chungnam National University

• Jiseon Kim, University of Washington - Seattle

• Ryoungsun Park, Wayne State University

• Hyeonjeong Jeon, Chungnam National University