论文标题:辅助信息在域估计中的应用模型与方法研究
Research on the Auxiliary Information Application to the Domain Estimation Models and Methodologies
论文作者
论文导师 刘建平,论文学位 硕士,论文专业 统计学
论文单位 暨南大学,点击次数 27,论文页数 64页File Size2572k
2005-05-01论文网 http://www.lw23.com/lunwen_378187112/
Auxiliary data; Application Model; Methodology; Domain Estimation
本文主要针对辅助信息在域估计中的应用模型与方法作了系统的阐述。文中对域估计方法的归纳是全新的,阐述的角度也是全新的。 首先,对本文的选题背景和研究意义以及国内外在这一研究领域的理论研究与实践的现状进行了概述。其次,界定了域、小域、小区域、辅助信息等基本概念,梳理阐明了辅助信息的来源、作用、局限性及其适用场合。第三,将各种域估计方法纳入到直接估计和间接估计两类方法的框架体系中。从辅助信息应用的角度,分别介绍了直接估计方法中的校准估计、广义回归估计和间接估计方法中的隐式模型估计和显式模型估计。在间接估计方法中,隐式模型估计除了介绍只使用行政记录和普查资料获得估计的传统人口统计学方法外,还介绍了需要样本辅助资料才能进行估计的合成估计法、组合估计法和James—Stein方法;显式模型估计方法中主要介绍了包含随机效应在内的线性混合模型,其中又分为区域层次的模型估计和单元层次的模型估计。第四,文章分别对直接估计方法和间接估计方法的使用条件和优缺点进行了小结;对隐式模型和显式模型估计方法进行了对比分析;以及对区域层次模型估计和单元层次模型估计的应用场合进行了比较。第五,文章介绍了模型参数估计的几种方法,包括经验最佳线性无偏估计法(EBLUP)、经验贝叶斯法(EB)和多层贝叶斯法(HB)。最后,对论文进行了总结,提出了本文的不足之处,并且对文章进行了理论和应用展望。
The paper systematically dissert various auxiliary information application models and methods in domain estimation. The system is new, and the point of view is also new.First, the paper summarizes the background of this topic, the purport of this research, and the theory and practice status quo of this field. Then gives some definitions such as domain, small domain, small area, auxiliary information, and introduces the sources, functions, applicability conditions and limitations of the auxiliary information. Thirdly, classifies the methods of domain estimation into two types: direct estimation method and indirect estimation method. Moreover, the latter method can be classified into implicit model estimate method and the explicit model estimate method. The article emphasizes the function of the auxiliary information in the domain estimation. Based on this purpose, the author separately discusses some direct estimation methods such as Calibration Estimation Method, Generalized Regression Estimation Method etc. and some indirect estimation methods such as Demographic Method, Synthetic Estimation Method, Composite Estimation Method, James-Stein Method (or Shrinkage Method) etc. which are all implicit model estimate methods, and some explicit model estimate methods, include area level linear mixed model and unit level linear mixed model. Fourthly, this paper gives some brief summary and contrast analysis of the direct estimation method and indirect estimation method, the implicit model and explicit model, and the area level model estimation and the unit level model estimation. Fifthly, paper gives some parameter estimation methods such as the Empirical Best Linear Unbiased Prediction method (EBLUP), Empirical Bayes method (EB) and Hierarchical Bayes method (HB). Finally, the author sum up some problems, points some shortcomings and prospects the theory development and applications of these methods.
|
