加速失效时间模型下现状数据的Jackknife模型平均  

The Jackknife Model Averaging of Accelerated Failure Time Model with Current Status

作　　者：赵慧 刘斌霞 董庆凯 张新雨[2] ZHAO HUI;LIU BINXIA;DONG QINGKAI;ZHANG XINYU(College of Statistics and Mathematics,Zhongnan University of Economics and Law,Wuhan 430070,China;Academy of Mathematics and Systems Science,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]中南财经政法大学统计与数学学院,武汉430070 [2]中国科学院数学与系统科学研究院,北京100190

出　　处：《应用数学学报》2023年第3期313-328,共16页Acta Mathematicae Applicatae Sinica

基　　金：国家自然科学基金(批准号:12171483)资助项目。

摘　　要：文章研究了响应变量为现状数据的情况下,加速失效时间模型的Jackknife模型平均方法.首先对数据进行合理的无偏变换,进而得到回归参数的最小二乘估计.然后引入删一交叉验证准则来选取候选模型的权重,并在一定正则性条件下,建立对应模型平均估计量的渐近最优性.此外,数值模拟表明,与现有的其他模型平均和模型选择方法相比,本文所提出的方法在预测上表现更佳.最后将所提方法应用于尼日利亚儿童死亡率的数据进行实证研究,进一步验证了所提方法的优良性质.This paper studies the jackknife model averaging method of the accelerated failure time model with current status data.Firstly,through the unbiased transfor-mation,we can obtain the LSE of regression parameters.Then the delete-one cross validation criterion is introduced to select the weights of candidate models,and under some regularity conditions,the asymptotic optimality of the model averaging estimator is established.Numerous simulation results show that the proposed method is superior to other existing model averaging and model selection methods in terms of prediction performance.Finally,we applied the proposed method to the NDHS data,and the real data also verify the excellent properties of the proposed method.

关 键 词：现状数据 加速失效时间模型 无偏变换 模型平均 渐近最优 

分 类 号：O212.1[理学—概率论与数理统计]