Model averaging for linear mixed models via augmented Lagrangian

Abstract

Model selection for linear mixed models has been a focus of recent research in statistics. Yet, the method of model averaging has been sparsely explored in this context. A weight finding criterion for model averaging of linear mixed models is introduced, as well as its implementation for the programming language R. Since the optimization of the underlying criterion is non-trivial, a fast and robust implementation of the augmented Lagrangian optimization technique is employed. Furthermore, the influence of the weight finding criterion on the resulting model averaging estimator is illustrated through simulation studies and two applications based on real data.