Poster #35 - The Impact of Nonnormal Residuals in Cross-Classified Multiple-Membership Random Effects Modeling

• In Event: 77.059 - MTCC Tuesday Poster Session 18
  In Poster Session: 77.059-4 - Educational Statisticians SIG Poster Session: Statistical Issues and Investigations

Abstract

This simulation study investigates how residual non-normality affects cross-classified multiple membership random effects modeling estimates. The conditions manipulated in the current study were: residual distribution type (normal, uniform and chi-squared), intra-unit correlation coefficient (0.1, 0.2 and 0.3), number of groups (20, 50 and 100), average group size (10 and 20), cross-classification rate (20% and 40%), and multiple membership rate (10% and 20%). The level-two cross-classification multiple membership factor random variance components were substantially biased with relatively fewer groups. In general, the degrees of relative bias were larger when the level-two residuals followed a chi-squared distribution (i.e., a severely skewed distribution) as compared to a uniform distribution (i.e., a non-normal but symmetrical distribution).

Authors

• Hyewon Chung, Chungnam National University

• Ryoungsun Park, SiFive

• Jiseon Kim

• Soyoung Park, Chungnam National University

• Hyeonjeong Jeon, Korean Educational Development Institute