统计研究 ›› 2019, Vol. 36 ›› Issue (5): 100-119.doi: 10.19343/j.cnki.11-1302/c.2019.05.008

• • 上一篇    下一篇

基于超总体模型的设计效应分解

徐映梅 杨延飞   

  • 出版日期:2019-05-25 发布日期:2019-05-25

Decomposition of Design Effects based on Super-population Models

Xu Yingmei & Yang Yanfei   

  • Online:2019-05-25 Published:2019-05-25

摘要: 本文基于超总体模型研究抽样调查中设计效应的计算问题。首先以随机效应模型为基础,明确了简单随机、二阶段、不等概率和分层抽样对应的超总体模型,进而通过所给模型推导出分层、类集、加权单因素设计效应的计算公式和多因素组合的设计效应计算公式并给出了对应估计量,公式表明:多因素同时存在的组合设计效应等于对应单因素设计效应的乘积。最后,对设计效应的理论值、估计值和真实值之间的关系进行了蒙特卡洛仿真,并利用相对偏倚、相对均方误进行了评价。本文的研究,对复杂抽样设计中正确计算、使用设计效应具有指导意义。

关键词: 设计效应, 超总体模型, 类集效应, 加权效应, 组合效应

Abstract: This paper studies the calculation of Design Effects in the sample surveys based on the super-population model. Firstly, taking the random effect model as a basis, the study clarifies the super-population models corresponding to SRS, two-stage sampling, unequal probability sampling and stratified sampling. Furthermore, through these Super-population Models, the formulas of the design effects of simple factors such as stratification, clustering and weighting,and the formulas of the combined design effects of multi-factors are derived with the corresponding estimators. The formulas show that the combined design effects of multiple factors is equal to the product of the design effect of a simple factor. Finally, the Monte Carlo simulation is conducted on the relationship between the theoretical value, the estimated value and the true value of the design effects, and their relative bias and relative mean square errors are evaluated. This research is of a guiding significance for the correct calculation and the application of the design effects in the complicated sampling.

Key words: Design Effect, Super-population Model, Clustering Effect, Weighting Effect, Combination effect.