Robust Estimation of the Variance Components in Semiparametric Linear Mixed Model

For a partial linear mixed model, we usually focus on the estimation of the variance components, and a lot of methods can be applied. However, many of these methods, such as maximum likelihood method and restricted maximum likelihood method, can be included in the framework of generalized estimating equation (GEE). As well known, the GEE method is sensitive to outliers. So, an alterative set of robust GEEs for both mean components and correlation parameters are proposed for the partial linear mixed model for longitudinal data in this paper. Some simulations are conducted to evaluate the performance of the proposed estimators. In the end, the method is illustrated with analysis of two real data sets.