Weighted Quantile Regression For Varying-Coefficient Models With Missing Covariates Based On Empirical Likelihood

The Research on missing data is becoming increasingly popular,the method of processing missing data is also gradually abundant.This paper mainly introduced the types and methods of processing missing data,different from the traditional way to deal with missing data,this paper innovatively proposed a parameter estimation method of varying-coefficient quantile regression model with partial missing covariates.In this paper,we proposed the inverse probability weighted estimation and the empirical likelihood weighted estimation of varying-coefficient quantile regression model on partial missing covariates.Before comparing the two estimates,we introduced the local weighted kernel estimation and the empirical likelihood estimation of varying-coefficient quantile regression model with complete data.Then we compared the inverse probability weighted estimation and the empirical likelihood weighted estimation with the missing data.From the asymptotic variance,the efficiency of the empirical likelihood weighted estimation is higher than that inverse probability weighted estimation,through the simulation,the performance of the probability of the above two estimates under the finite sample was further evaluated.On the basis of the varying-coefficient quantile regression model based on the empirical likelihood,this paper also made a supplementary study and introduction to the former method.The level of consumption data of Jilin Province as an example,studied the varying-coefficient regression estimation,using examples to compare the fitting effect of regression coefficient and linear regression,for the final fitting performance,variable-coefficient regression model fitting effect was more accurate and wider.Finally,discussed and studied several interval estimation methods for quantile regression models,and compared the confidence interval length and coverage rate.It is concluded that different quantile regression models are more effective under different sample conditions.