摘要
在右删失数据下,当删失指标随机缺失时,对条件分布函数分别构造了校准加权核估计,插值加权核估计以及逆概率加权核估计;然后由这些估计分别导出了条件分位数的核估计,并建立了这些估计的渐近正态性;最后,在有限样本下,对这些估计进行了数值模拟,分析了各估计的优缺点.
In this paper,when the censoring indicators of right-censored data are missing at random,a weighted calibration kernel estimator,a weighted imputation kernel estimator and a weighted inverse probability kernel estimator of conditional distribution function are investigated,respectively.Then,the kernel estimators of conditional quantiles are derived from these estimators,and the asymptotic normality of these estimators is established.Finally,the estimators are numerically simulated under finite samples,and the advantages and disadvantages of each estimator are analyzed.
作者
周杨程
王江峰
袁汶汶
张惠利
ZHOU Yang-cheng;WANG Jiang-feng;YUAN Wen-wen;ZHANG Hui-li(School of Statis.Math.,Zhejiang Gongshang Univ.,Hangzhou 310018,China)
出处
《高校应用数学学报（A辑）》
北大核心
2020年第4期379-392,共14页
Applied Mathematics A Journal of Chinese Universities(Ser.A)
基金
浙江省自然科学基金(LY18A010007)
国家社会科学基金(20BTJ049)
浙江省一流学科A类(浙江工商大学统计学)。
关键词
条件分位数
加权核估计
渐近正态性
删失指标
随机缺失
conditional quantile
weighted kernel estimation
asymptotic normality
censoring indicator
missing at random
