The Impact of Prior Distributions for Variance Parameters in Cross-Classified Multiple-Membership Random Effects Modeling

• In Event: 18.047 - Statistical Research in the Contexts of Propensity Scores, Longitudinal Models, and Bayesian Analyses

Abstract

This simulation study examines how the selection of priors affects variance parameter estimates in cross-classified multiple membership random effects modeling. Two prior distributions were used: uniform distribution and inverse gamma distribution with four different scale factors (1, 1), (.01, .01), (.001, .001), and (1, .001). Among the five priors, inverse-gamma distribution (.001, .001) resulted in the smallest relative bias. Substantial over-estimation of the level two variance component was identified for both uniform distribution and inverse-gamma distribution (1, 1). With a larger number of level two units, the degree of relative bias was diminished. The results suggest that researchers should be cautious in selecting accurate priors, especially with a relatively small number of level two units.

Authors

• Hyewon Chung, Chungnam National University

• Ryoungsun Park, Wayne State University

• Jiseon Kim

• Hyeonjeong Jeon, Chungnam National University