Efficient Shrinkage for Generalized Linear Mixed Models Under Linear Restrictions

Abstract

In this paper, we consider the pretest, shrinkage, and penalty estimation procedures for generalized linear mixed models when it is conjectured that some of the regression parameters are restricted to a linear subspace. We develop the statistical properties of the pretest and shrinkage estimation methods, which include asymptotic distributional biases and risks. We show that the pretest and shrinkage estimators have a significantly higher relative efficiency than the classical estimator. Furthermore, we consider the penalty estimator LASSO (Least Absolute Shrinkage and Selection Operator), and numerically compare its relative performance with that of the other estimators. A series of Monte Carlo simulation experiments are conducted with different combinations of inactive predictors, and the performance of each estimator is evaluated in terms of the simulated mean squared error. The study shows that the shrinkage and pretest estimators are comparable to the LASSO estimator when the number of inactive predictors in the model is relatively large. The estimators under consideration are applied to a real data set to illustrate the usefulness of the procedures in practice.