Efficient imputation and resampling methods for analyzing complex survey data

加拿大统计局等组织经常进行大规模复杂的调查，以收集有关目标人口的关键信息。调查数据的统计分析为管理战略、发展计划和政策的明智决策提供了依据，这些决策影响着我们社会的各个方面。调查数据也被研究人员广泛用于解决商业、教育、卫生、行为以及其他社会心理和经济领域的问题。在复杂调查数据分析的挑战性问题中，缺失数据问题和方差估计技术一直是方法论研究的两个重要领域。该建议扩展了当前实践中使用的方法，并开发了新的有效的输入方法来处理缺失数据和有效的重采样程序来进行方差估计。我们提出的方法建立在目前使用的程序的基础上，为一些问题提供了解决方案，并可能改变公共使用调查数据集的产生方式和这些数据集的分析方式。提出的研究课题也为研究生提供了充分的机会，让他们参与并将想法发展到论文研究中。

Large scale complex surveys are routinely conducted by organizations such as Statistics Canada to gather critical information about the target population. Statistical analysis of survey data provides grounds for informed decisions on management strategies, development plans and policies that affect every aspects of our society. Survey data are also widely used by researchers to address issues in business, education, health, behaviour and other psychosocial and economical areas. Among the challenging issues related to analysis of complex survey data, missing data problems and variance estimation techniques have been two important areas for methodological research. This proposal extends methods that are used in current practice and develops new and efficient imputation methods for handling missing data and efficient resampling procedures for variance estimation. Our proposed methods build on the strength of currently used procedures, provide solutions to some of the problems and will likely change the way that public-use survey data sets are produced and how these data sets are analyzed. The proposed research topics also provide ample opportunities for graduate students to get involved and to develop ideas into their thesis research.