Statistical Inference Of Semiparametric Regression Models With Nonignorable Missing Data

Missing data are commonly encountered in biomedicine,epidemiology,social science,economics,psychology,among others.Most of existing methods are assumed that the missingness data mechanism is ignorable.However,in many practical applications,the missing data are nonignorable.Therefore,this paper studies statistical inference of semiparametric regression models with nonignorable missing data.This study is not only the needs of the development of practical problems,and the needs of the development of statistics.Its research has important value.This paper discusses the estimation problem of parameters and nonparametric function in semiparametric regression models with nonignorable missing data.First,we construct a Logistic regression model for nonignorable missing data mechanism.We use B-splines to transform nonparametric function to regression model.Then,we use MCMC algorithm to estimate the parameters and nonparametric function.Finally,the availability of this method is proved by simulation.This paper considers Bayesian variable selection based on semiparametric regression model with nonignorable missing data.Bayes-Lasso and Bayes-ALasso are used to simultaneously select variables and estimate parameters.Simulation studies are conducted to investigate the performance of the proposed Bayesian variable select method.An example is used to illustrate the application of Bayesian variable selection.