病例队列设计下可加可乘风险模型的Bootstrap估计  

Bootstrap Estimation for General Additive-Multiplicative Hazards Model under Case-Cohort Design

作　　者：陈佳琪 CHEN Jia-qi(Institute of Applied Mathematics,Academy of Mathematics and Systems Science,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院数学与系统科学研究院,应用数学研究所,北京100190 [2]中国科学院大学,北京100049

出　　处：《数理统计与管理》2025年第1期106-118,共13页Journal of Applied Statistics and Management

摘　　要：本文聚焦于病例队列设计的研究,旨在解决协变量测量成本高昂和疾病发病率较低等现实问题。该方法可以显著降低大规模预测性研究的成本。针对病例队列数据下的可加可乘风险模型,本文提出了三种估计方法,并采用了一种单阶段重抽样的非参数bootstrap方法进行协方差估计。通过数值模拟研究,我们将bootstrap方法获得的协方差估计与基于渐近理论的协方差估计进行了比较。结果显示,bootstrap方法表现出色。此外,我们将所提方法应用到一个实际案例中,取得了良好的分析效果。Our study focuses on the case-cohort design as a cost-effective alternative to large prospective cohort studies,effectively addressing challenges such as expensive covariate measurements and rare events.We propose three estimation methods for a general additive-multiplicative hazards model with case-cohort data,and use a nonparametric bootstrap method with a single-stage resampling to estimate variance.Through simulation studies,we compare variance estimates obtained by the bootstrap method with asymptotic theoretical estimates.The bootstrap method demonstrates strong performance.Furthermore,we provide a real-life example to illustrate the application of our proposed method.

关 键 词：可加可乘风险模型 BOOTSTRAP方法 病例队列设计 简单随机抽样 单阶段重抽样 加权估计函数 

分 类 号：O212[理学—概率论与数理统计]