非概率样本模型辅助校准估计方法  

A Model-assisted Calibration Estimation Method for Non-probability Samples

作　　者：潘莹丽 刘展 Pan Yingli;Liu Zhan(Farulty of Mathematics and Statistics,Hubei University,Wuhan 430062,China;Huhei Key Laboratory of Applied Mathematics,Hubei University,Wuhan 430062,China)

机构地区：[1]湖北大学数学与统计学学院,武汉430062 [2]湖北大学应用数学湖北省重点实验室,武汉430062

出　　处：《统计与决策》2021年第22期5-10,共6页Statistics & Decision

基　　金：国家自然科学基金资助项目(11901175)。

摘　　要：大数据时代,如何高效地从高维数据中挖掘有效信息并进行统计推断逐渐成为人们关注的热点问题。文章基于高维数据研究如何充分利用已知辅助信息对非概率样本进行校准,主要包括两个方面的内容:一是在总体规模已知的情况下,对非概率样本进行模型辅助SCAD校准和模型辅助ALASSO校准,依次求出校准权重并对总体均值进行统计推断。二是在总体规模未知的情况下,对非概率样本进行估计控制模型辅助SCAD校准和估计控制模型辅助ALASSO校准,依次求出校准权重并对总体均值进行统计推断。In the era of big data,how to efficiently mine effective information from high-dimensional data and make statistical inference has gradually become a hot issue.This paper is based on high-dimensional data to study how to make full use of known auxiliary information to calibrate non-probability samples.It mainly includes two aspects:Firstly,the model-assisted SCAD calibration and the model-assisted ALASSO calibration are used for non-probability samples when the overall population is known,and then the calibration weight is calculated in turn and the statistical inference is carried out for the population mean.Secondly,when the overall population is unknown,the estimated control model-assisted calibration with SCAD and the estimated control model-assisted calibration with ALASSO are used for non-probability samples,and then the calibration weight is calculated in turn and the population mean is statistically inferred.

关 键 词：非概率样本 模型辅助校准 估计控制模型辅助校准 SCAD ALASSO 

分 类 号：C811[社会学—统计学]