基于分层设计效应的样本分配研究  

Research of Sample Allocation Based on Stratification Design Effect

作　　者：罗薇[1,2] 

机构地区：[1]广东工业大学管理学院,广州510006 [2]广东工业大学大数据战略研究院,广州510006

出　　处：《统计与决策》2017年第16期18-21,共4页Statistics & Decision

基　　金：国家社会科学基金资助项目(17BTJ037);广东省哲学社会科学"十二五"规划学科共建项目(GD15XGL09)

摘　　要：文章从分层抽样的角度,以设计效应为工具,构造出若干种分层抽样的设计效应模型,应用于分层抽样的样本分配设计。由于分层设计效应主要取决于各层平均数的差异,则差异越大,分层的效果越好。为了满足目标变量全国估计量和特定小域估计量的精度要求,样本在各小域的最优分配可能是比例分配和非比例分配的折中。From the perspective of stratified sampling the paper takes design effect as a tool to construct several stratification design effect models, which are applied to sample allocation. As the stratification design effect is dominated by the difference of each stratum average, the greater the difference is, the better the stratification effect is. In order to satisfy the accuracy requirement of national estimates and estimates for various regions of the country, the optimal allocation of samples in each small area is likely to he the compromise of proportional allocation and non-proportional allocation.

关 键 词：分层抽样 设计效应 样本分配 

分 类 号：C811[社会学—统计学]