Diagnostic Measures For Semiparametric Regression Model With Longitudinal Data Based On Empirical Likelihood Confidence Region

In this paper, we consider the diagnostic measures for semiparametric regression model with longitudinal data based on empirical likelihood confidence region. Statistics diagnostic study the rationality among the data, model and inference. Empirical likeli-hood is nonparametric method to construct confidence region. Two diagnostic statistics based on empirical likelihood are proposed:the ratio of confidence region volume,the sum of deviation squares because of the superiorities. One of the host study spots is longitu-dinal data, which can be explained well by semiparametric regression model. The third statistic is the difference value of regression function based on the characteristics of the semiparametric regression model. Simulation studies and a case analysis are conducted to compare the diagnostic effect of three proposed statistics.