Akaike's Information Criterion in Generalized Estimating Equations

TLDR

Correlated response data are common in biomedical studies, and generalized estimating equations (GEE) are increasingly used, but few model‑selection criteria exist for GEE because the standard Akaike Information Criterion (AIC) relies on maximum likelihood, which GEE does not use. The study proposes a modified AIC that replaces likelihood with quasi‑likelihood and adjusts the penalty term for GEE. The authors evaluate the modified AIC via simulation studies and illustrate its application to a real dataset. Summary.

Abstract

Summary. Correlated response data are common in biomedical studies. Regression analysis based on the generalized estimating equations (GEE) is an increasingly important method for such data. However, there seem to be few model‐selection criteria available in GEE. The well‐known Akaike Information Criterion (AIC) cannot be directly applied since AIC is based on maximum likelihood estimation while GEE is nonlikelihood based. We propose a modification to AIC, where the likelihood is replaced by the quasi‐likelihood and a proper adjustment is made for the penalty term. Its performance is investigated through simulation studies. For illustration, the method is applied to a real data set.