Weighted Local Linear Dual Kernel Estimation Of Conditional Quantile Under Right-censored Dependent Data

Survival data is composed of survival time and covariate factors that affect survival time.Due to the different integrity of information,it is divided into complete data and censored data.Censored data is more common in life and has great research value.Among them,the diversification of the types of censored data affects the diversification of research methods.From linear regression models to quantile regression models,from parametric to nonparametric estimation methods,all are to better fit complex data types and data performance in real life.In the case that mean regression cannot meet the research needs,the idea of quantile regression and non-parametric methods are introduced,and methods such as kernel estimation,local polynomial,and local linear estimation are compared in conditional quantile estimation,It is found that local linear estimation can not only effectively alleviate boundary problems,but is also an estimation method that can be applied to more data types.At present,it is difficult to study weighted local linear estimation of conditional distribution function,conditional density function and conditional quantile in independent samples.It is more practical and valuable to explore in the data environment of deleting dependent data and multivariate covariance.Therefore,the purpose of this study is to use an optimized local linear estimation method to construct conditional quantile estimators in the case of right-censored dependent data and multivariate covariates.In the process of research,we first construct the weighted local linear dual-kernel estimation of the conditional density function,and then derive the weighted local linear dual-kernel estimation of the conditional quantile,establish and prove the asymptotic normality of these estimators.Finally,With a limited sample,the nature of the estimation is studied through computer numerical simulation research and an actual data about heart failure patients,and the offset,standard deviation and overall mean square error at a fixed point are compared,which reflects the weighted local linear dual-core estimation method,and it is highly adaptable when applied to right-censored dependent data.