A Comparison of Methods for Centering Covariates in Cross-Classified Random Effects Models

• In Event: Hierarchical Linear Models

Abstract

The cross-classified random-effects model (CCREM) is used to handle cross-classified data in which units are nested within multiple higher-level factors. One of the assumptions when estimating a CCREM is that the covariates do not correlate with the random effects at any level. When handling nested data by estimating a multilevel model, covariate centering (cluster-mean centering) offers an approach to reduce the impact of assumption violations. In this study, we propose extending cluster-mean centering that partitions the within-cluster and between-cluster effects of a level-1 regression coefficient for use with a CCREM. We conducted a Monte Carlo Simulation study and demonstrated that centering alternatives provide less biased and precise estimates than covariate coefficient estimation for a CCREM without centering.

Authors

• Young Ri Lee, University of Texas at Austin
  Presenting Author

• Tasha Beretvas, University of Texas at Austin
  Non-Presenting Author

• James E. Pustejovsky, University of Wisconsin - Madison
  Non-Presenting Author