Combining information from independent surveys, small area estimation of complex parameters, analysis of complex survey data

抽样调查以概率抽样为基础，被广泛用于提供关于各种重要主题的可靠当前统计数据。使用可变性度量(标准误差和可信区间)来评估样本统计的质量。调查越来越多地被用来提供小范围的统计数据，如县对贫困学龄儿童的估计，这是制定政策和方案、分配资金和营销决策所需的。结合来自独立调查(数据融合)的信息，以获得有效的估计，并研究并非在同一调查中观察到的变量之间的关系，已受到极大的关注。调查数据被社会和健康科学家和其他人用来研究关系和检验感兴趣的假设，例如，来自加拿大国家公共卫生调查的纵向数据。我的研究计划将集中在所有上述当前感兴趣的话题上。我尤其计划研究以下问题：(A)我在小区域估计方面做了大量工作，我2003年出版的Wiley一书已成为研究人员和用户关于基于模型的方法的标准参考，即使在区域内样本量较小的情况下，这种方法也可以导致可靠的估计。我建议继续开展基于模型的方法方面的工作，特别是利用现实的模型和最佳估计方法，估计复杂的小区域参数，如世界银行在发展中国家使用的贫困衡量标准。(B)多层次模型经常用于分析来自多阶段抽样的数据，例如学校内的学校样本(2级)和学生样本(1级)。我计划研究我最近开发的一种新方法的理论性质和扩展，该方法可以得出有效的推断，并提供一种统一的方法。(C)我最近关于综合独立调查资料的工作提供了估算总数的有效方法。将研究扩展以处理更复杂的参数，如回归系数和丢失的数据。

Sample surveys, based on probability sampling, are widely used to provide reliable current statistics on a variety of important topics. Measures of variability (standard errors and confidence intervals) are used to assess the quality of the sample statistics. Surveys are increasingly being used to provide small area statistics, such as county estimates of poor school-age children, that are needed in formulating policies and programs, allocation of funds and marketing decisions. Combining information from independent surveys (data fusion) to obtain efficent estimates and study relationships between varaibles not all observed in the same survey has received much attention. Survey data are used by social and health scientists and others to study realtionships and to test hypotheses of interest, e.g, longitudinal data from the National Public Health Survey of Canada. My research program will focus on all the above topics of current interest. In particular, I plan to study the following problems among others:(a) I have done extensive work on small area estimation and my 2003 Wiley book has become a standard reference to researchers and users on model-based methods that can lead to reliable estimates even when the sample sizes are small within areas. I propose to continue my work on model-based methods, especially for estimating complex small area parameters such as poverty measures used by the World Bank in developing countries, using realistic models and optimal estimation methods. (b) Multi-level models are often used to analyse data from multistage sampling, e.g. sample of schools (level 2) and students (level 1) within schools. I plan to study the theoretical properties and extensions of a new method I have recently developed that leads to valid inferences and provides a unified approach. (c) My recent work on combining information from independent surveys provides efficient methods of estimating totals. Extensions to handle more complex parameters such as regression coefficeints, and missing data will be studied.