Semiparametric Analysis Based On Right Censored Data For Transformation Models With Missing Covarites

Missing covariate data are very common in regression analysis.There are many reasons for this matter,and there are also many ways to solve it.In this paper,the classical method of inverse probability weighted estimators is used.Under missing at random,we choose the transformation model,and obtain two scoring equations by likelihood and martingale.The weighting of the covarites are calculated by kernel estimation.Moreover,according to different initial values,sample size,censoring rate and specific models,we conducted four sets of numerical simulation experiments.Finally,a real dataset is analyzed to illustrate the proposed methods.The main contents are organized as follow:Chapter Two: Basic knowledge of some theory.First,we introduce the concept of survival analysis,including type of data and some common functions.Then,we talk about three classic missing mechanism,and we discuss some methods to solve it.Next,we introduce the counting process,inverse probability weighted and kernel estimation.Chapter Three: Introduction of transformation models and parameter estimates.Chapter Four: Numerical studies.Considering the methods on missing covarites,we change different initial values,sample size,censoring rate and specific models on simulations,and some conclusions have also been verified.Chapter Five: Case study.Considering the right censored,the inverse probability weighted estimators is used to real dataset.