Abstract:
Mixed models are commonly used for analyzing clustered data, including
longitudinal data and repeated measurements. Unrestricted full maximum
likelihood (ML) methods have been extensively studied in the literature
for analyzing generalized, linear, and mixed models. However, constraints
or parameter orderings may occur in practice, and in such cases, we can
improve the efficiency of a statistical method by incorporating parameter
constraints into the ML estimation and hypothesis testing. In this talk, I
will discuss constrained inference with generalized linear mixed models
(GLMMs) under linear inequality constraints. Methods will be assessed
using both Monte Carlo simulations and actual survey data from a health
study.