Huangya Li, Qianqian Zhu, Maozai Tian

【Abstract】

Based on quantile regression and nonparametric model, this paper studies the estimation methods and algorithms of the nonparametric quantile regression model under censored data. We apply the redistribution-of-mass idea to extend the estimation method of parametric model to nonparametric area, and get a locally weighted estimator. Meanwhile,in order to make full use of the data, we combine the idea of the inverse probability weights and redistributionof-mass weights, applying both of the weights simultaneously to non-censored data and censored data to obtain a new locally combined weighted estimator. Furthermore, through Monte Carlo simulation, conclusions are drawn that under fixed censoring,the redistribution-of-mass weighted estimator outperforms the others, while under random censoring, the combined weighted estimator is more efficient and reasonable. Finally, the estimation methods are illustrated on real data, and the results further demonstrate the efficiency and reasonableness of the proposed estimation methods.

【Keywords】

quantile regression, nonparametric model, censored data, inverse probability weights, redistribution-of-mass