A $C_p$ type criterion for model selection in the GEE method when both scale and correlation parameters are unknown

Abstract

In this paper, we consider a model selection criterion using the GEE method including unknown scale and correlation parameters. We propose a model selection criterion for selecting variables and a working correlation structure. Under some regularity conditions, we showed that our criterion is the same as the criterion proposed by Inatsu and Imori [8]. A numerical study reveals that we can reduce the prediction error by selecting both variables and a working correlation structure.