Consistent selection of working correlation structure in GEE analysis based on Stein’s loss function

Abstract

In this paper, we attempt to select a working correlation structure for generalized estimating equations. We propose a selection criterion based on the Stein’s loss function. Our criterion consistently selects the true correlation structure when the unknown parameters are $sqrt(n)$-consistent, where $n$ is the sample size. We demonstrate the performance of the proposed methodology by a numerical study.