"Nonparametric Estimation with Applications to Large and Complex Survey Data"

该奖项是根据2009年《美国复苏和再投资法案》(公法111-5)提供资金的。今天的调查统计学家面临的一个困难是调查的结构越来越复杂。美国拥有各种各样的纵向调查，与广泛使用的横断面数据相比，这些调查在捕捉动态人口关系方面具有相当大的优势。从这些复杂的调查中做出尽可能无模型的推论是可取的。非参数统计是一种灵活且前景看好的工具，可以恰当地反映复杂的设计结构。然而，同时考虑高维调查误差的检测、平滑以及复杂相关结构带来的额外复杂性，给非参数调查分析带来了巨大的挑战。研究人员致力于新的非参数模型辅助方法，用于大型和复杂的调查，包括纵向调查，通过纳入“廉价的”辅助信息。目前的项目包括(1)开发更精细、更智能的调查抽样工具；(2)研究存在非抽样误差(如无响应和测量误差)的非参数调查方法；(3)探索纵向调查分析的新程序和新理论。随着许多研究领域的大型复杂调查越来越多地为公众所用，本方案中的理论和实践可以作为调查从业者、(生物)统计学家、流行病学家、经济学家、社会学家等研究人员的重要工具，所提出的方法将极大地丰富纵向调查建模的技术，并拓宽对调查抽样的传统理解。这项拟议的研究还将通过向几个联邦机构(包括人口普查局、国家卫生统计中心、司法统计局和劳工统计局)的调查研究人员提供现代和先进的调查方法，来加强美国联邦统计系统。

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5). One difficulty facing today's survey statisticians is the increasingly complex structures of surveys. The U.S. is very well provided with various sorts of longitudinal surveys which have considerable advantages over widely used cross-sectional data for capturing dynamic demographic relationships. It is desirable to make inferences from these complex surveys as model-free as they can be. Nonparametric statistics is a flexible and promising tool that properly reflects complex design structures. However, the simultaneous consideration of detection of survey errors with high dimensionality, smoothing and the additional complexity emerging from complex correlation structures presents great challenges in nonparametric survey analysis. The investigator works on novel nonparametric model-assisted methods for large and complex surveys, including longitudinal surveys, via incorporation of "cheap" auxiliary information. The current project includes (1) developing finer and more intelligent nonparametric tools for survey sampling; (2) investigating nonparametric survey methodology in the presence of nonsampling errors, such as nonresponse and measurement errors;(3) exploring new procedures and novel theory in longitudinal survey analysis.The field of survey research is undergoing profound and rapid changes brought on by larger societal, technological, and theoretical developments. With large complex surveys in many research areas becoming increasingly available for public use, the theory and practice in this proposal can serve as an important tool for survey practitioners, (bio)statisticians, epidemiologists, economists, sociologists, and other researchers.The proposed methodologies will significantly enrich the techniques of longitudinal survey modeling and broaden the traditional understanding of survey sampling. The proposed research will also strengthen the U.S. federal statistical system by providing survey researchers from several federal agencies (including Census Bureau, the National Center for Health Statistics, the Bureau of Justice Statistics, and the Bureau of Labor Statistics) modern and advanced methods in survey methodology.