Adaptive Fitting of Linear Mixed Effects Models with Correlated Random Effects

• Activity Number: 523
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Guangxiang George Zhang* and John J. Chen+
• Companies: State University of New York at Stony Brook and State University of New York at Stony Brook
• Keywords: Longitudinal data analysis ; mixed-effects ; nonconvergence rate ; condition number ; collinearity ; transformation

Abstract

Mixed-effect models (MEMs) have been widely used in longitudinal data analysis as they allow for correlations among repeated measurements from the same unit. How to best model random effects of MEMs is still an important and unresolved issue in practice. We propose a data-driven algorithm to adaptively fit the MEMs that reduces the correlation among random effects in transformed parameter space. Simulations show that the proposed algorithm significantly improves the non-convergence rate, the reduction of correlation among random effects and the fitted log-likelihoods. Two real data sets are used to illustrate the application of this algorithm.