Generalized Varying Coefficient Partially Linear Models For Clustered Data

摘要

We are considering estimation in a generalized varying coefficient partially linear model for clustered data using generalized estimating equations.This article is proposing an EM algorithm technique to estimate the parametric component and the nonparametric coefficient function.From simulation we can see clearly that,in order to get the general consistent estimator of the parametric part,we need not undersmoothing the nonparametric part.Unlike other method in estimating the parameter and the nonoparametric coefficient function,the optimal bandwidth for the model is similar to nonparametric model,so the method of leaving-one-cluster-out cross-validation could be directly used in selecting bandwidth.