Performance of Covariate-Based Partitioning Goodness of Fit Test for Semiparametric Logistic GEE Regression

Abstract

This paper evaluates the covariate-based partitioning goodness of fit (GOF) test for semiparameric logistic regression for correlated binary data. Estimation of the model uses GEESmoothing Spline, where the basis of estimation is GEE and the estimation of nonparametric component is based on smoothing spline. In this paper we extend the covariate-based partitioning GOF test for parametric logistic GEE model into GOF test for semiparametric logistic GEE model. The performance of this extension method is evaluated by simulation. We obtained that it has good capability to detect correct model but low power to detect incorrect model.