Modelling of Covariance Structures in Generalized Estimating Equations for Longitudinal Data

Mis-speci cation of covariance structure; Modelling of mean-covariance structures Mathematical Subject Classification: 62J05 Abstract: When modelling longitudinal data the technique of generalized estimating equations allows for speci-cation of a working covariance structure that is intended to approximate the true covariance structure of repeated measures. Mis-specication of the working covariance structure, however, may lead to a great loss of estimate ef-ciency for mean parameters. Based on a modied Cholesky decomposition of covariance matrices, in this talk we propose a data-driven approach to model the mean and covariance structures in the generalized estimating equations, simultaneously. The proposed approach has no assumption of distribution but assumes the rst four moments of responses exist. The resulted estimates for mean and covariance parameters are shown to be consistent and asymp-totically Normally distributed. Real data analysis and simulation studies show that the proposed approach yield