改进样本代表性的多目标追加平衡设计  

A Balanced Design with Added-up Multi-purpose Sample to Improve Sample Representatives

作　　者：巩红禹[1,2] 陈雅 Gong Hongyu;Chen Ya

机构地区：[1]内蒙古财经大学 [2]内蒙古财经大学经济应用统计评价研究中心

出　　处：《统计研究》2018年第12期113-122,共10页Statistical Research

基　　金：国家社会科学基金项目"平衡抽样设计及其在政府统计调查中的应用研究"(15BTJ008);国家社会科学基金项目"农村危房改造的增长和福利效应的大数据动态SAM分析"(17BTJ024);内蒙古2018年青年科技英才支持计划项目(NJYT-18-A17)的资助

摘　　要：本文主要讨论样本代表性的改进和多目标调查问题。本文提出了一种增加样本量与调整样本结构相结合的方法——追加样本的平衡设计,即通过追加样本,使得补充的样本与原来的样本组合生成新的平衡样本,相对于初始样本,减少了样本与总体的结构性偏差;平衡样本通过选择与多个目标参数相关的辅助变量,使得一套样本对不同的目标参数而言都具有良好的代表性,进而完成多目标调查。结合2010年第六次全国人口普查分县数据,通过选择多个目标参数,对追加样本后的平衡样本进行事后评估的结果表明,追加平衡设计能够有效改进样本结构,使得样本结构与总体结构相近,降低目标估计的误差;同时也说明平衡抽样设计能够实现多目标调查,提高样本的使用效率。This paper mainly deals with two issues, i.e., improvement of sample representatives and multi-purpose sample survey. This paper proposes a new multi-purpose sampling methodology to improve sample representatives, an integration of added-up sample size and adjustment of sample composition, i.e., to create a new balanced sample from the original sample integrated with the added-up sample. In comparison with the original sample, the new integrated sample has reduced the composition bias between the original sample and the population. The balanced sample refers to Horvitz Thompson Estimator based on total auxiliary variables equal to the real value of the population total. Furthermore, a set of balanced samples can offer good representatives for various purposive parameters by selecting auxiliary variables related to the multi-purpose parameters and perform well for the multi-purpose sample survey. Based on the balanced sample with added-up using census data by county from the Sixth Population Census in 2010, a post facto assessment using multi-purpose parameters shows that the balanced design with add-up sample can improve sample composition in an effective way, make the sample and the population compositions close enough and reduces the errors in the purposive estimation. At the same time, the balanced sample design is justified for the multi-purpose survey with enhanced sample efficiency.

关 键 词：代表性样本 追加样本 平衡样本 多目标 

分 类 号：O212[理学—概率论与数理统计]