带流数据集AFT模型的可再生估计  

A Renewable Estimation of AFT Model With Streaming Data Sets

作　　者：潘莹丽 黄河[3] Pan Yingli;Huang He(Faculty of Mathematics and Statistics,Hubei University,Wuhan 430062,China;Hubei Key Laboratory of Applied Mathematics,Hubei University,Wuhan 430062,China;Faculty of Management,Wuzhou University,Wuzhou Guangxi 543002,China)

机构地区：[1]湖北大学数学与统计学学院,武汉430062 [2]湖北大学应用数学湖北省重点实验室,武汉430062 [3]梧州学院管理学院,广西梧州543002

出　　处：《统计与决策》2023年第9期47-52,共6页Statistics & Decision

基　　金：国家自然科学基金资助项目(11901175)。

摘　　要：流式生存数据是一个随时间延续而无限增长的动态生存数据集合,由于数据集以流的形式不断高速到达,一旦当前批次的数据到来,经过快速处理后就要被释放,不能继续保留在内存中。基于右删失流式生存数据来解析协变量与生存时间之间的相关性时,加速失效时间模型(AFT模型)是常被使用的模型之一。文章基于带流数据集的AFT模型,通过泰勒展开构造一个Working估计方程,提出可再生估计,该估计仅依赖历史批数据集的汇总统计量和当前批数据集,有效避免了计算机对历史批数据存储带来的压力。模拟分析和实证结果表明,基于带流数据集的AFT模型提出的可再生估计方法在有限样本中的运行性能较好,在实践中具有可操作性。Streaming survival data is a dynamic survival data set that grows infinitely over time.As the data set is constantly arriving at high speed in the form of stream,once the current batch of data arrives,it will be released after rapid processing and cannot be retained in memory.The accelerated failure time(AFT)model is one of the most commonly used models to analyze the correlation between covariates and survival time based on right censored streaming survival data.Based on AFT model with streaming data set,this paper constructs the Working estimation equation through Taylor expansion,and proposes a renewable es⁃timation.This estimate only relies on the aggregate statistics of historical batch data sets and the current batch data set,effective⁃ly avoiding the pressure brought to the computer by the historical batch data store.Simulation analysis and empirical results show that the renewable estimation method based on AFT model with streaming data sets has better performance in limited samples and is operable in practice.

关 键 词：流数据集 右删失 AFT模型 可再生估计 

分 类 号：F222[经济管理—国民经济] O212[理学—概率论与数理统计]