A Comparison of the Three Bayes Point Estimates in Bayesian Quadratic Cross-Classified Multiple Membership Growth Curve Modeling

• In Event: Statistical Methods for Complex Multilevel Structures and Data Types

Abstract

The Bayesian quadratic cross-classified multiple membership growth curve modeling has been introduced to fit a non-linear growth curve as well as to handle non-pure multilevel dataset in which repeated measures are nested within students, cross-classified by first school attended and a set of subsequent schools. This study compared parameter recovery across the three Bayes point estimates (i.e., the posterior mean, median, and mode) under a variety of simulated conditions. Overall, it would be recommended for researchers to report the posterior mode estimates over posterior mean or median estimates for school variance component except some conditions and to use a minimum of more than 200 schools with 50 students in order to avoid substantial biased estimation of school level variance component.

Authors

• Hyewon Chung, Chungnam National University

• Soyoung Park, Chungnam National University

• Hyeonjeong Jeon, National Youth Policy Institute

• Jiseon Kim