Misspecification in Generalized Linear Mixed Models with Correlated Random Effects Based on Randomization

• Activity Number: 37
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Mohammad Zakir Hossain* and Heiko Grossmann and Steven G. Gilmour
• Companies: University of London and University of London and University of Southampton
• Keywords: Correlation structure ; Random effects distribution

Abstract

We investigate the effect of misspecification of the correlation structure and the random effects distribution on the estimated fixed effects parameters and variance components in a generalized linear mixed model. The variance-covariance matrix of the random effects is derived from randomization ideas. We present the results of simulation studies where the misspecified model with uncorrelated random effects is fitted to data generated from the model with correlated random effects. We also fit the model with normally distributed random effects to data simulated from models with different random effects distributions. We find that misspecification of the correlation structure and the distribution of random effects has hardly any effect on estimates of fixed effects parameters but estimates of variance components are frequently severely biased.