高维数据下平均处理效应估计的调整方法

Adjusting Method Effect of Average Processing Effect Estimate in High-dimensional Data

平均处理效应的估计一直是统计和经济领域研究的热点问题,低维空间下处理效应的估计量在高维数据下表现往往欠佳,文章结合变量选择与分层抽样,在高维数据下研究了平均处理效应估计的调整方法,讨论了调整后估计量的无偏性和渐近性,建立了估计量精度与回归系数估计精度之间的关系,结果表明,分层回归调整可以控制影响处理效应的变异源,从而提高估计的准确性。最后借助随机模拟进一步验证了结论的可靠性。

The estimation of average processing effect has always been a hot topic in the field of statistics and economics,and estimators of processing effects in low-dimensional space tend to perform poorly in high-dimensional data.Combined with variable selection and stratified sampling,this paper studies the adjusting method of average processing effect estimation in high-dimensional data,discusses the unbiased and asymptotic properties of the adjusted estimators,and also establishes the relationship between the estimator accuracy and the regression coefficient estimation accuracy.The results show that stratified regression adjustment can control the variation sources that affect the treatment effect,thus improving the accuracy of estimation.Finally,the paper verifies the reliability of the conclusion by use of stochastic simulation.