Statistical Inference Of Partial Linear Modal Models With Missing Data

In the field of practical application,the data distribution is often asymmetrical and biased.In this case.the partial linear modal model is an important method to characterize these features.Because the partial linear modal model is the overall improvement of the partial linear model and the modal regression model.On the one hand,some linear models retain the characteristics of non-parametric smoothness and overcome the curse of dimension.On the other hand,mode broadens the scope of the regression,especially when the data obeys the heavy tail distribution or has abnormal values,compared with the mean regression,the modal regression can provide a more meaningful prediction and obtain better results.However,most of the studies on partial linear modal models are carried out under complete data.But in practical studies,due to the negligence of respondents or involving sensitivity problems,a large number of data will be missing.For the missing data,there is no systematic research on the parameter estimation of partial linear modal models.Therefore,this paper mainly studies the statistical inference of partial linear modal models with missing data.This paper is divided into four parts.The first chapter introduces the significance and background of partial linear modal model.In the second chapter,the kernel estimation methods of partial linear modal model and non-parametric part are briefl,y introduced,and the parameters of the model are estimated with MEM algorithm.In the third chapter,a kind of inverse probability weighted idea is introduced to deal with the problem of missing covariant by using the mode-weighted MEM algorithm.In the fourth chapter,the method of single interpolation,multiple interpolation and mode interpolation is used to deal with the missing of response variables,and then the MEM algorithm is used to estimate the parameters of the model.