统计与决策2025，Vol.41Issue(18)：5-10,6.DOI:10.13546/j.cnki.tjyjc.2025.18.001
数据缺失条件下基于模型平均调整的因果效应估计方法
Causal Effect Estimation Method Based on Model Average Adjustment Under the Condition of Missing Data
摘要
Abstract
When observational data is used to make causal inference,confounding variable bias and missing covariate data will reduce the estimation accuracy of processing effects.In order to solve these problems,this paper proposes a method of estimat-ing causal effect based on model average adjustment for data with missing covariates.In the proposed method,the missing data is first interpolated by using the interpolation method,then the model average adjustment method is employed to perform weighted average on multiple propensity score estimation models,and the advantages of each model are synthesized.Finally,the reliability and accuracy of propensity score estimation are improved by dual adjustment mechanisms.The experimental results show that compared with the inverse probability weighted method based on logistic regression,the proposed method can effectively reduce the impact of confounding bias and covariate data missing,and improve the estimation accuracy of ATE,thus providing a new way to deal with the causal effect estimation of covariate missing data.关键词
协变量缺失/因果效应/逆概率加权/模型平均Key words
missing covariates/causal effect/inverse probability weighting/model average分类
数理科学引用本文复制引用
耿智琳,张丽丽,张耀峰,张志刚..数据缺失条件下基于模型平均调整的因果效应估计方法[J].统计与决策,2025,41(18):5-10,6.基金项目
国家社会科学基金重点项目(23ATJ005) （23ATJ005）
国家社会科学基金资助项目(19BTJ030) （19BTJ030）
湖北省教育厅科学研究计划项目(D20222202) （D20222202）
