抽样调查数据分析中权数引入问题研究  

Research on the Introduction of Weights in Sampling Survey Data Analysis

作　　者：刘晓宇 武雅萱 金勇进[2,3] Liu Xiaoyu;Wu Yaxuan;Jin Yongjin(School of Statistics,Capital University of Economics and Business,Beijing 100070,China;School of Statistics,Renmin University of China,Beijing 100872,China;Center for Applied Statistics,Renmin University of China,Beijing 100872,China)

机构地区：[1]首都经济贸易大学统计学院,北京100070 [2]中国人民大学统计学院,北京100872 [3]中国人民大学应用统计科学研究中心,北京100872

出　　处：《统计与决策》2023年第8期18-24,共7页Statistics & Decision

基　　金：首都经济贸易大学新入职青年教师科研启动基金资助项目(XRZ2023076)。

摘　　要：抽样调查中的权数可以放大样本信息,使得样本体现出的特征能反映总体情况。在许多社会学研究中,例如用抽样调查数据进行回归分析,经常不考虑权数。虽然有时是不必引入权数的,但现实中大多数情况缺乏对该问题的思考。是否将权数引入推断中,需要依据对权数质量的评估进行判断。采用质量不高的权数会造成估计精度与估计效率的降低。文章针对现有方法中人为影响较大、检验假设可能失效等缺陷,提出采用非参数方法对目标变量及其加权值进行一致性检验的思路,以解决权数在抽样调查数据分析中的引入问题。The weights in the sampling survey can enlarge the sample information,and the characteristics of the sample can reflect the characteristics of the population.In many sociological studies,such as regression analysis using survey data,weights are often not considered.Although sometimes it is unnecessary to include weights,in reality,most cases lack the consideration of this problem.Whether to introduce weight into inference needs to be judged according to the evaluation of weight quality.Using weights with low quality will reduce the accuracy and efficiency.In view of the defects of existing methods such as large artificial influence and possible failure of test assumptions,this paper proposes the idea of using non-parametric methods to test the consistency of target variables and their weighted values to solve the problem of introducing weights in the analysis of sample survey data.

关 键 词：抽样推断 总体代表性 权数质量评估 一致性检验 

分 类 号：C811[社会学—统计学]