复杂抽样设计下的域估计问题研究

发布时间：2018-12-20 17:05

【摘要】：随着国内定量研究方法的开展和大型调查数据的免费公布,研究者不仅使用抽样调查数据对总体分析,还需要对域总体进行分析。本文对调查数据满足域精度推断的域估计问题进行研究。首先,根据实际调查中的域估计问题,指出解决域估计问题最好的方法是事先确定好需要估计的域,并在抽样设计时兼顾域的估计精度。但是,在实际调查中还包含计划外的域,通过对简单随机抽样下的域估计问题的研究,说明非计划域的估计问题的最大难点是域样本量的随机性。然后,针对实际中的抽样调查数据多来源于分层、多阶段、整群和不等概率等复杂抽样设计的问题,指出需要结合复杂抽样设计信息、域样本量的随机性、域样本在总体的误差层和误差群中的分布,对复杂抽样设计下的域估计问题进行研究。最后,以中国家庭追踪调查(China Family Panel Studies,CFPS)为例,对复杂抽样设计下的域估计问题进行案例研究。
[Abstract]:With the development of domestic quantitative research methods and the free publication of large-scale survey data, researchers not only use sample survey data to analyze the population, but also need to analyze the field population. In this paper, the problem of domain estimation of survey data satisfying domain accuracy inference is studied. Firstly, according to the domain estimation problem in the actual investigation, it is pointed out that the best way to solve the domain estimation problem is to determine the domain which needs to be estimated in advance, and to take into account the estimation accuracy of the domain in the sampling design. However, the unplanned domain is also included in the actual investigation. By studying the domain estimation problem under the simple random sampling, it is shown that the biggest difficulty of the unplanned domain estimation problem is the randomness of the domain sample size. Then, aiming at the problem that the sampling data come from stratification, multi-stage, cluster and unequal probability, it is pointed out that it is necessary to combine the information of complex sampling design and the randomness of domain sample size. The problem of domain estimation under complex sampling design is studied based on the distribution of domain samples in the error layer and error group of the population. Finally, a case study of domain estimation under complex sampling design is carried out by taking (China Family Panel Studies,CFPS (China Household tracking Survey) as an example.
【作者单位】：北京大学中国社会科学调查中心;
【基金】：国家社会科学基金项目“追踪调查中小域估计的方法及其应用研究”(11CTJ005);国家社会科学基金项目“平衡抽样设计及其在政府统计调查中的应用研究”(5BTJ008)的资助国家自然科学基金项目“基于涵盖误差的我国周期性普查数据质量评估方法:理论与应用研究”(71301033)
【分类号】：C31

【参考文献】