【作者】 杨杰；

【导师】 丁洁丽；

【作者基本信息】 武汉大学 ， 概率论与数理统计， 2022， 硕士

【摘要】 在许多大型队列研究中,往往需要对大量研究个体进行长期追溯和随访,当重要影响因素或协变量的采集非常昂贵时,采用传统的简单随机抽样可能会导致试验过于昂贵而超过预算.因此,发展和研究能节约成本和提高效率的抽样机制具有非常重要的意义.对于带有删失的生存数据而言,病例队列设计是应用最为广泛的有偏抽样机制之一.此种抽样设计首先通过简单随机抽样方式从全队列中抽取一个子队列,子队列中的个体和子队列之外的所有病例个体组成病例队列样本.这种有偏抽样方式能有效地提高研究效率、节约研究成本.本文主要研究了如何应用加性风险模型拟合由病例队列设计获取的生存数据的问题.利用加权估计方法,通过模拟研究获得了这种方法在有限样本量下的优良表现,并评估了病例队列设计相较于简单随机抽样设计的有效性.进一步,推广此方法应用于两个实际数据中,展示其在实际中的应用价值和应用前景.本文的结构安排如下:第一章中,介绍本文的研究背景,综述发展现状,总结本文主要工作.第二章中,引入加性风险模型,介绍病例队列设计下参数的加权估计方法,得到估计的渐近性质.第三章中,通过统计模拟来评估估计方法在有限样本量下的表现.第四章中,分析两个实际数据,展示估计方法在实际应用中的优良表现.第五章中,对本文工作进行总结,对未来研究工作进行展望.更多还原

【Abstract】 Large cohort studies often require long-term retrospection and follow-up of a large number of research individuals.When the collection of important influencing factors or covariates is very expensive,simple random sampling may cause the experiment to be too expensive and exceed the budget.Therefore,it is significance to develop and research sampling mechanisms that can save costs and improve efficiency.For censored survival data,case cohort design is one of the most widely used biased sampling mechanisms.In this sampling design,firstly,a subcohort is selected from the full cohort through a simple random sampling method.The case-cohort sample consists of the subcohort and all case outside subcohort.This biased sampling method can save research costs and improve research efficiency.In this paper,we discuss how to fit the additive hazards model to case-cohort data.We consider a weighted estimating approach,and conduct simulation studies to evaluate the good finite-sample performance of the estimation method,and evaluate the efficiency of the case-cohort design to the simple random sampling scheme.Furthermore,we apply the proposed method to analyze two real datas.The structure of this paper is as follows:The first chapter introduces the research background,sums up the development of the research direction,and summarizes the main work of paper.The second chapter introduces the additive risk model and the weighted estimation method of parameters under the case cohort design and reviews its asymptotic theory.The third chapter conducts simulation studies to evaluate the good finite-sample performance of the estimation method.The fourth chapter apply the proposed method to analyze two real datas,and show the excellent performance of the estimation method in practical applications.The fifth chapter sums up the work of this paper,and prospects for future research work.更多还原