因果效应估计在SAS软件中的实现  

The Application of Causal Effect Estimation in SAS

作　　者：黄慧杰 张洪璐 于泽洋 宋德胜 王淳 刘媛媛[1] 崔壮[1] 李长平[1] Huang Huijie;Zhang Honglu;Yu Zeyang(Epidemiology and Biostatistics Institute,School of Public Health,Tianjin Medical University,300070,Tianjin)

机构地区：[1]天津医科大学公共卫生学院流行病与卫生统计学教研室,300070

出　　处：《中国卫生统计》2023年第1期9-14,共6页Chinese Journal of Health Statistics

基　　金：教育部人文社会科学研究项目(17YJAZH048,20YJAZH021);国家自然科学基金青年基金(81803333)。

摘　　要：目的利用SAS开发的CAUSALTRT过程,实现三类估计方法的因果效应估计。方法采用SmokingWeight数据集,以戒烟为处理变量,体重变化为结局变量,其他因素为混杂变量,通过增强逆概率加权法(augmented inverse probability weighting,AIPW)对平均处理效应(the average treatment effect,ATE)进行估计,通过回归调整法(regression adjustment,REGADJ)对处理组平均处理效应(the average treatment effect for the treated,ATT)进行估计。结果戒烟对体重变化的ATE和ATT分别为3.209(95%CI:2.232~4.187)和3.276(95%CI:2.332~4.219)。结论CAUSALTRT可以实现不同的因果效应估计,但应用时需要考虑其是否满足前提假设以及注意事项。Objective Using the CAUSALTRT procedure in SAS to realize the causal effect estimation by the three types of estimation methods.Methods Using the Smoking Weight dataset,taking smoking cessation as the treatment variable,weight change as the outcome variable,and other factors as confounding variables,the average treatment effect(ATE)was estimated by augmented inverse probability weighting(AIPW),and the average treatment effect for the treated was estimated by regression adjustment(REGADJ).Results The estimates for ATE and ATT of smoking cessation on weight change were 3.209(95%CI:2.232~4.187)and 3.276(95%CI:2.332~4.219).Conclusion CAUSALTRT can estimate different causal effects,but it needs to consider whether it meets the assumptions and precautions before using it.

关 键 词：因果效应估计 SAS 平均处理效应 CAUSALTRT 

分 类 号：R195.1[医药卫生—卫生统计学]