Collaborative Research: Computation-driven small area inference with applications

该奖项是根据2009年美国复苏和再投资法案（公法111-5）资助的。越来越需要在国家和国家以下各级评估人类福祉的各种措施（包括社会经济和健康特征）。然而，由于数据收集成本、自然限制、保密性和其他道德问题等制约因素，国家以下一级的数据可用性往往有限。本研究旨在从多个数据源中提取用于预测目的的相关信息，并在有限数据条件下提供可靠的预测。由于小区域统计数据通常由不同的利益相关者群体使用，因此迫切需要开发不同的可靠风险度量方法——这是本项目的主要重点。更具体地说，将开发一种统一的计算方法，以在广泛的统计模型、模型参数估计方法、数据类型以及对称或非对称损失函数中估计不同的风险措施。拟议的研究将取代复杂的分析公式，通常很难或不可能获得，通过简单有效的技术，利用高功率计算的效率。将使用计算机生成的复杂调查数据集对拟议的技术进行广泛评价。本研究将小区域统计的范围扩展到公共利益的几个应用，并将加强相关的教育计划。

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).There is a growing demand to assess various measures of human well-being (including socio-economic and health characteristics) at national and sub-national levels. However, data availability at the sub-national level is often limited owing to constraints such as costs of data collection, natural limitations, confidentiality and other ethical issues. This research aims to extract relevant information for predictive purposes from several data sources and provide sound predictions under limited-data conditions. Since small area statistics are routinely used by a diverse group of stakeholders, there is an urgent need to develop different reliable risk measures - this is the prime focus of this project. More specifically, a unified computational method will be developed to estimate different risk measures across a broad spectrum of statistical models, methods of estimation of model parameters, types of data, and symmetric or asymmetric loss functions. The proposed research will replace complicated analytical formulae, often hard or impossible to obtain, by simple efficient techniques that use the efficiencies of high power computing. Extensive evaluation of the proposed techniques will be carried out using computer-generated and complex survey datasets. This research will extend the scope of small area statistics to several applications of public interest and will strengthen related educational programs.