Comparing the Three Bayes Point Estimates in Multiple Membership Multilevel Modeling

• In Event: Bayesian Estimation and Propensity Score Analysis in Educational Statistics

Abstract

This simulation study was designed to evaluate the parameter recovery in Bayesian multiple membership multilevel modeling. To test the accuracy of multiple membership multilevel modeling parameter estimation, three Bayes point estimates via Markov chain Monte Carlo (MCMC) were compared: the posterior mean, the posterior median, and the posterior mode. The posterior mode estimates of level two variance components were substantially under-estimated with 50 groups or less. However, the posterior mean estimates of level two variance components were substantially over-estimated with 30 groups. In general, the use of posterior median estimates resulted in less biased parameter estimates as compared to the use of posterior mean or mode estimates. The coverage rates of 95% credible intervals were close to nominal coverage.

Authors

• Hyewon Chung, Chungnam National University

• Ryoungsun Park, Wayne State University

• Jiseon Kim, University of Washington - Seattle

• Hyeonjeong Jeon, Chungnam National University