多阶连续抽样设计下广义加权回归估计方法研究  

The Study of Generalized Weighted Regression Estimation Methods Based on Multistage Successive Sampling Design

作　　者：陈光慧[1] CHEN Guang-hui(School of Economics,Jinan University,Guangdong Guangzhou 510632,China)

机构地区：[1]暨南大学经济学院

出　　处：《数理统计与管理》2019年第6期996-1004,共9页Journal of Applied Statistics and Management

基　　金：国家社会科学基金项目(18BTJ005);国家“万人计划”青年拔尖人才支持项目(W02070337)的阶段性成果

摘　　要：在抽样调查领域中,关于抽样方案设计的研究应用较为充分和完整,但关于抽样估计的研究应用却较为缺乏和滞后。本文首先总结了国外相关研究成果,研究了基于广义加权回归的抽样估计方法,同时证明其满足渐近设计无偏和最小化渐近期望方差的理论条件。同时,本文以各类常见的抽样设计为基础,通过模型组和模型水平将现有的超总体回归模型进行扩展,基于复杂的多阶连续抽样调查,建立各种类型的超总体回归模型进行模型辅助的广义加权回归抽样估计,给出了具体的回归估计步骤和结果,最终形成一套关于广义加权回归抽样估计的理论方法体系,为抽样估计方法在我国政府统计部门中的有效应用奠定理论基础。In the field of sampling survey,the research and application of sampling scheme design is more full and complete,but the research and application of sampling estimation is relatively short and lagging.On the basis of summarizing the related research results of foreign countries,this paper improves and extends a set of generalized weighted regression sampling estimation methods,and gives the theoretical conditions to satisfy asymptotically design unbiased and minimized asymptotically expected variance.At the same time,based on the common sampling design,the existing super-population regression model is extended through the model group and the model level.Based on the complex multistage successive sampling survey,various types of super-population regression models are set up,then the generalized weighted regression sampling estimation are used,based on the model-assisted.The steps and results are estimated,and a set of theoretical method system for the generalized weighted regression sampling estimation is formed,which lays a theoretical foundation for the effective application of sampling estimation method in the Government Statistics Department of China.

关 键 词：抽样估计 超总体模型 广义加权回归估计量 多阶连续抽样 

分 类 号：C811[社会学—统计学] O212[理学—概率论与数理统计]