Comparing Bayes Point Estimates in Cross-Classified Multiple Membership Random Effects Modeling With Nonnormal Residuals

• In Event: Educational Statisticians SIG Paper Session

Abstract

This simulation study was designed to evaluate the parameter recovery in Bayesian cross-classified multiple membership random effects modeling. Three Bayes point estimates were compared: the posterior mean, the posterior median, and the posterior mode under the variety of manipulated conditions such as non-normality of residuals (Chi-square distribution of one degree of freedom and uniform), priors (inverse-gamma distribution (.001, .001) uniform distribution), number of groups (20, 30, 50 and 100), average group size (10, 20 and 40), cross-classification rate (20% and 40%), and multiple membership rate (10% and 20%). In general, with the inversed gamma prior, the MCMC posterior median is recommended for estimating cross-classified multiple membership random effects modeling with non-normal residuals.

Author

• Hyewon Chung, Chungnam National University