基于广义加乘危险率模型的多元数据病例队列试验研究.docVIP

基于广义加乘危险率模型的多元数据病例队 列试验研究 郑明 ，孙怡 ，郁文 复旦大学管理学院统计学系，上海　200433 摘要：在大型流行病学跟踪研究中，病例队列抽样设计是一种常用的降低抽样成本的抽样方 法。在这种抽样设计中，只要求观测所有病例样本以及非病例样本中一个容量较小的随机子样 本的协变量。现有文献对于一元数据的病例队列设计有较多研究，但对于多元数据的研究较 少。本文基于广义加乘危险率模型，用拟得分方程方法对于多元数据的病例队列设计进行参数 估计，考虑两种随机子样的抽样机制，讨论了所得估计量的渐近性质，并用模拟计算和一个实 例来展示所提出估计量的表现。 关键词：病例队列设计，加乘危险率模型，多重事件数据，聚类数据 中图分类号： O212 Multivariate additive-multiplicative hazard models for case-cohort studies ZHENG Ming , SUN Yi , YU Wen Department of Statistics, School of Management, Fudan University, Shanghai 200433 Abstract: Case-cohort design is commonly applied in practice to reduce cost in large epidemiological studies especially when the event rate is low and the cost of covariate assembling for all subjects is huge. Under such a design, people only collect covariate histories for the cases and a small random subset selected from the cohort. This design is extensively explored for univariate failure time data under various survival models. However, literatures on case-cohort analysis for multivariate failure time data are limited and are only based on the proportional hazards model. In this article, we concentrate on the case-cohort analysis with multivariate failure time data under a class of marginal additive-multiplicative hazard models. Two sampling schemes for the subcohort, including Bernoulli sampling and strati?ed simple random sampling, are discussed. Estimation procedures are developed and we explore the asymptotic properties of the resulting estimators. Some simulation studies are conducted to assess the ?nite sample performances of the proposed inferential procedures. A real example is also provided for illustration. 基金项目： 高等学校博士学科点专项科研基金新教师类资助课题 (项目编号：20110071120023), 国家自然科学基金资助课题 (项目编号11101091). 作者简介： Zheng Ming, female, professor, major research direction：survival analysis, sampling survey. Sun Yi, female, master candidate. Correspondence author：Yu Wen, male, assistant professor, major research direction: semiparametric model, survival analysis,