无论是传统的医学统计，还是现代的大数据分析，大规模队列研究都非常普遍和必须，但此类队列研究常常需要花费大量的金钱与时间来测量所有研究个体的自变量信息。两阶段队列抽样方法可以大幅降低抽样规模，减少抽样成本。常用的两阶段队列抽样包括病例-队列抽样、病例-对照抽样以及巢式病例-对照抽样。现有文献对于这三种抽样设计在各种模型假设下的统计推断以及基于数据分析技术的效率提升进行了广泛地研究，而在不增加抽样规模的前提下，从抽样设计本身出发来提升统计推断效率的研究却鲜有人为之。本项目提出一类不同于传统设计的新的两阶段队列抽样设计思想，针对若干种不同类型的生存数据设计与之相适应的抽样方法，并在一些常用的生存分析半参数模型假设下，结合似然与逆概率加权技术为新抽样设计所获得的数据提供效率更高的统计推断方法。新抽样方法效率的提高来源于抽样设计本身，其应用空间很大，本项目的研究将进一步丰富大规模数据处理的分析手段。

In both traditional medical statistics and modern big data analysis, large cohort studies often appear and receive much attention. Usually it is prohibitively expensive and time consuming to collect covariates information on all the subjects involved in such large cohorts. Thus, two-phase cohort sampling designs are used to significantly reduce the cost on data collection. Commonly used two-phase cohort sampling designs include case-cohort sampling, case-control sampling and nested case-control sampling. The statistical inference methods with these three designs are extensively studied by the existing literature under various models, and some efficiency improving approaches based on data analysis techniques are also provided. However, little existing literature has tried to improve the estimating efficiency with comparable sample size through the sampling designs in the second phase. In this project, we propose a new idea for the two-phase cohort sampling designs, which could be more efficient than the existing ones. Based on the proposed idea, we give out proper sampling designs for different kinds of survival data. Moreover, the likelihood method and the inverse probability weighing approach are applied to analyze the data drawn by the new sampling design under several commonly used semiparametric survival models. The proposed idea improves the efficiency in terms of the design itself, and has promising potential in application. The research of the project complements the existing statistical analysis techniques for large scale data.