Residual Normality Assumption and the Estimation of Multiple Membership Random Effects Models

• In Event: Issues in Parameter Estimation in Complex Multilevel Models

Abstract

Fitting a multiple membership random effects model (MMrem) to non-purely nested data can account for lower-level unit interdependencies and the contextual effects of multiple higher-level units on the outcomes of lower-level units. To extend research on the influence of residual non-normality from purely hierarchical to multiple membership data, this Monte Carlo simulation study examined two-level MMrem fixed and random effect parameter estimate biases under a fully-crossed design. Preliminary results showed that estimates of fixed and level-one random effect parameter estimates are not sensitive to level-two residual non-normality, whereas small level-two sample size was associated with biased level-two random effect parameter estimates. The coverage rate of the 95% credible interval consistently deviated from the nominal value when level-two residuals were non-normal.

Authors

• Jieru Chen, Georgia State University

• Audrey J. Leroux, Georgia State University