Statistical Inference In Varying Coefficient Partially Nonlinear Models With Missing Response Variable At Random

Semiparametric regression model in the study of statistics is essential.This kind of model compared with the parameter model and nonparametric regression model has stronger explanatory and adaptability.This paper mainly consider varying coefficient partially nonlinear model that is one of important semiparametric regression model,in which the linear part is replaced with nonlinear function of covariates.this model has more flexibility.In practical problems,the data we collected are often missing.The existing methods for dealing with complete case data are no longer applicable,and the methods of inverse probability weighting and imputation are commonly used to study missing data.In this paper,we mainly study the estimation of varying coefficient partially nonlinear model when the response variable are missing at random.We use the profile least-squares and inverse probability weighting imputation method to estimate the regression coefficient of parametric and nonparametric function,in which the propensity score is estimated by the covariate balancing propensity score method.the estimators of nonlinear parameters are shown to be asymptotically normal.Simulation studies show that the estimators has robust properties,and we also give the real data analysis.