摘要
对于调查数据来说,合理有效地实现权数的构建与调整,是提高调查数据推论准确度的重要方法之一。由于调查数据通常会存在几类影响调查数据代表性的问题,本研究从权数构建与调整的角度提出有针对性的解决方案:其一,对于概率抽样调查,就调查设计、实施、完成过程中出现的不等概率、无应答、覆盖率不足等问题,提出包括广义回归在内的几种权数构建和调整方法;其二,对于非概率抽样调查中的选择性偏差问题,着重探讨如何用倾向性得分的方法来实现样本权数的构建和应用。
Whether sample weights can be reasonably and effectively constructed and adjusted is one of the important aspects to improve the inference accuracy in using survey data.Subject to problems affecting the representativeness of survey data,this study puts forward targeted solutions from the perspective of weight construction and adjustment.On the one hand,for probability sampling survey,this study puts forward several weight construction and adjustment methods including generalized regression method to solve the problems of unequal probability,no response and insufficient coverage in the process of survey design,implementation and completion.On the other hand,for non-probability sampling survey that has severe selection bias problem,this paper focuses on how to use propensity score method to realize the construction and application of sample weight.
作者
邹宇春
李建栋
ZOU Yuchun;LI Jiandong
出处
《湖南师范大学社会科学学报》
CSSCI
北大核心
2022年第1期50-59,共10页
Journal of Social Science of Hunan Normal University
关键词
权数
概率抽样
非概率抽样
weight
probability sampling
non-probability sampling