《工程科学学报》录用稿，htps:/doi.org/10.13374/i,issn2095-9389.2021.07.04.002©北京科技大学2020 《工程科学学报》编辑部 因果推断三种分析框架及其应用综述1 马忠贵，徐晓晗，刘雪儿 北京科技大学计算机与通信工程学院，北京100083 ☒通信作者，E-ail:g_runeko@163.com 摘要探寻事物之间的因果效应在统计学、计算机科学、计量经济学等许多领域都是一个焕具力的研究课题。伴 随着2021年约书亚·安格里斯特(Joshua D.Angrist)和圭多·因本斯(GuidoW.Imbens)因对因果关系分析的 方法学贡献”而获得诺贝尔经济学奖，因果推断必将在这些领域大放异彩。本文简要众绍闲果推断所涉及的基本概 念及其三种分析框架：反事实框架、潜在结果模型和结构因果模型。首先，从反事实椎架介绍因果效应的发端：然后， 从基于反事实的两个因果推断分析框架：潜在结果模型和结构因果模型，来分别御述网个分析框架所涉及的关键理 论和应用方法。其中，潜在结果模型使用数学和可计算的语言对因果理论进行阐述、是一种将假设、命题和结论清晰 化表达的计算模型，其在原因和结果变量已知的前提下定量分析原因变量对结果变量的因果效应，并对缺失的潜在 结果进行补齐，使观察性研究的效果接近试验性研究。结构因果模型则是一种基于图论的因果推断方法，它将事件 分为观察、干预和反事实三个层级，并通过o运算将干预和反事著层级的因果关系都降维成可以通过统计学手段解 决的问题。最后，本文探讨了现今多领域内因果推断的应用场景 并总结了三种分析框架的异同点。 关键词因果效应：因果推断：反事实：潜在结果模型：结构医 分类号TG142.71 Three Analytical l Frameworksòf Causal Inference and Their Applications MA Zhong-gui,XU Xiao-han回，LIU Xue-er School of Computer and Communication Engineering. University of Science and Technology Beijing,Beijing 100083,P.R. China Corresponding author, 0163.com ABSTRACT Causality is a generic relationship between an effect and the cause that produces it,the causal relationship among things is a problem we have been thinking about,but the complexity of causality is sometimes far beyond our imagination.Even if some causality problems seem easy to analyze,it may not be easy to get an exact answer.However, through the continuous innovation and development of empirical research methods in recent decades,we have had several clear analytical frameworks and effective methods for how to define causality and determine the degree of causality. Exploring the causal effects among things is a promising research topic in many fields such as statistics,computer science, econometrics,etc.With Joshua D.Angrist and Guido W.Imbens winning the Nobel Prize in economics for their methodological contributions to the analysis of causality in 2021,causal inference will shine in these fields.This article briefly introduces the basic concepts involved in causal inference,and its three analytical frameworks:Counterfactual Framework(CF),Potential Outcomes Framework(POF)and Structural Causal Model(SCM).Firstly,we introduce the origin of causal effects according to CF.Secondly,based on the counterfactual theory,there are two analysis frameworks,called POF and SCM,and we introduce the key theories and methods respectively.The SCM explains the causal theory through 败离日期：2021-07-04 演自中央高校基本科研业务费专项资金资助项目RE-DF-20-12,ERF-GF-18-0ZB) 地址：北京市海淀区学院路30号 邮政编码：100083 电话：010-62333436 E-mail:xuebaozr@ustb.edu.cn http://cje.ustb.edu.cn《工程科学学报》编辑部 因果推断三种分析框架及其应用综述1 马忠贵，徐晓晗，刘雪儿 北京科技大学计算机与通信工程学院，北京 100083  通信作者，E-mail: g_runeko@163.com 摘 要 探寻事物之间的因果效应在统计学、计算机科学、计量经济学等许多领域都是一个颇具潜力的研究课题。伴 随着 2021 年约书亚·安格里斯特（Joshua D. Angrist）和圭多·因本斯(Guido W.Imbens)因“对因果关系分析的 方法学贡献”而获得诺贝尔经济学奖，因果推断必将在这些领域大放异彩。本文简要介绍因果推断所涉及的基本概 念及其三种分析框架：反事实框架、潜在结果模型和结构因果模型。首先，从反事实框架介绍因果效应的发端；然后， 从基于反事实的两个因果推断分析框架：潜在结果模型和结构因果模型，来分别阐述两个分析框架所涉及的关键理 论和应用方法。其中，潜在结果模型使用数学和可计算的语言对因果理论进行阐述，是一种将假设、命题和结论清晰 化表达的计算模型，其在原因和结果变量已知的前提下定量分析原因变量对结果变量的因果效应，并对缺失的潜在 结果进行补齐，使观察性研究的效果接近试验性研究。结构因果模型则是一种基于图论的因果推断方法，它将事件 分为观察、干预和反事实三个层级，并通过 do 运算将干预和反事实层级的因果关系都降维成可以通过统计学手段解 决的问题。最后，本文探讨了现今多领域内因果推断的应用场景，并总结了三种分析框架的异同点。 关键词 因果效应；因果推断；反事实；潜在结果模型；结构因果模型 分类号 TG142.71 Three Analytical Frameworks of Causal Inference and Their Applications MA Zhong-gui, XU Xiao-han , LIU Xue-er School of Computer and Communication Engineering, University of Science and Technology Beijing, Beijing 100083, P. R. China Corresponding author, E-mail: g_runeko@163.com ABSTRACT Causality is a generic relationship between an effect and the cause that produces it, the causal relationship among things is a problem we have been thinking about, but the complexity of causality is sometimes far beyond our imagination. Even if some causality problems seem easy to analyze, it may not be easy to get an exact answer. However, through the continuous innovation and development of empirical research methods in recent decades, we have had several clear analytical frameworks and effective methods for how to define causality and determine the degree of causality. Exploring the causal effects among things is a promising research topic in many fields such as statistics, computer science, econometrics, etc. With Joshua D. Angrist and Guido W. Imbens winning the Nobel Prize in economics for their methodological contributions to the analysis of causality in 2021, causal inference will shine in these fields. This article briefly introduces the basic concepts involved in causal inference, and its three analytical frameworks: Counterfactual Framework(CF), Potential Outcomes Framework(POF) and Structural Causal Model(SCM). Firstly, we introduce the origin of causal effects according to CF. Secondly, based on the counterfactual theory, there are two analysis frameworks, called POF and SCM, and we introduce the key theories and methods respectively. The SCM explains the causal theory through 1收稿日期:2021-07-04 基金项目: 中央高校基本科研业务费专项资金资助项目(FRF-DF-20-12, FRF-GF-18-017B) 地址：北京市海淀区学院路 30 号 邮政编码：100083 电话：01062333436 E-mail: xuebaozr@ustb.edu.cn http://cje.ustb.edu.cn 《工程科学学报》录用稿，https://doi.org/10.13374/j.issn2095-9389.2021.07.04.002 ©北京科技大学 2020 录用稿件，非最终出版稿