哈佛大学Miguel Hernan科学家最新2019年《因果推断:概念与方法》书稿终版,280页讲解因果效应(附下载)

1 月 3 日 专知

【导读】哈佛大学公共卫生学院(HSPH)Miguel Hernan与Jamie Robins 教授共同编著了关于因果逻辑推断方面的书作《因果推断:概念与方法》,总共分3个部分,21章,280多页,对因果推理的概念和方法做了系统性阐述,是各个领域包括经济学、健康医疗、心理学、计算机等从业人士的重要参鉴材料。



Hernán MA, Robins JM (2019). Causal Inference. Boca Raton: Chapman & Hall/CRC, forthcoming.

Miguel Hernan教授因果推断书稿下载

 请关注专知公众号(扫一扫最下面专知二维码,或者点击上方蓝色专知

  • 后台回复“CI2019” 就可以获取《因果推断:概念与方法》的下载链接~ 

  • 专知2019年1月将开设《深度学习:算法到实战》课程,欢迎关注报名!

       专知开课啦!《深度学习: 算法到实战》, 中科院博士为你讲授!


简介

我(Miguel Hernan)和同事杰米·罗宾斯(Jamie Robins)正在写一本书,书中对因果推理的概念和方法进行了连贯的介绍。目前,这些材料大多分散在几个学科的期刊上,或者局限于技术文章。我们希望这本书能引起任何对因果推理感兴趣的人的兴趣,例如流行病学家、统计学家、心理学家、经济学家、社会学家、政治学家、计算机科学家……这本书分为三个难度越来越大的部分:没有模型的因果推理、有模型的因果推理和复杂纵向数据的因果推理。



目录

Introduction: Towards less casual causal inferences v

I Causal inference without models 1

1 Adefinition of causal effect 3

1.1 Individual causal effects. . . . 3

1.2 Average causal effects . . . 4

1.3 Measures of causal effect. . . 7

1.4 Random variability . . . 8

1.5 Causation versus association . . . . 10

2 Randomized experiments 13

2.1 Randomization . . . 13

2.2 Conditional randomization  . . . . 17

2.3 Standardization . . . . . 19

2.4 Inverse probability weighting . . 20

3 Observational studies 25

3.1 Identifiability conditions  . . 25

3.2 Exchangeability . . 27

3.3 Positivity . . . . 30

3.4 Consistency: First, define the counterfactual outcome . . 31

3.5 Consistency: Second, link to the data . . . 35

3.6 The target trial . . 36

4 Effect modification 41

4.1 Definition of effect modification  . . 41

4.2 Stratification to identify effect modification . . . . 43

4.3 Why care about effect modification. . . . 45

4.4 Stratification as a formof adjustment . . . 47

4.5Matching as another formof adjustment . 49

4.6 Effect modification and adjustmentmethods. . 50

5 Interaction 55

5.1 Interaction requires a joint intervention. . . . 55

5.2 Identifying interaction . . . . 56

5.3 Counterfactual response types and interaction. . . . 58

5.4 Sufficient causes  . . . . . . 60

5.5 Sufficient cause interaction. . 63

5.6 Counterfactuals or sufficient-component causes?. . . . 65

6 Graphical representation of causal effects 69

6.1 Causal diagrams . . . . 69

6.2 Causal diagrams andmarginal independence . . . . 72

6.3 Causal diagrams and conditional independence . . . 73

6.4 Positivity and well-defined interventions in causal diagrams . . 75

6.5 A structural classification of bias. . . . . . 79

6.6 The structure of effect modification  . . . . . . 80

7 Confounding 83

7.1 The structure of confounding .  . . 83

7.2 Confounding and exchangeability . . . . 85

7.3 Confounders . .  . . . 86

7.4 Single-world intervention graphs. . . . . 91

7.5 How to adjust for confounding  . . . 93

8 Selection bias 97

8.1 The structure of selection bias  . . . . 97

8.2 Examples of selection bias. . . . . . . . 99

8.3 Selection bias and confounding  . 101

8.4 Selection bias and censoring .. . . . . . . 103

8.5 How to adjust for selection bias . . . . . . . . 105

8.6 Selection without bias  . . . 108

9 Measurement bias 111

9.1Measurement error . . . . . 111

9.2 The structure ofmeasurement error. . . . . 112

9.3 Mismeasured confounders . . . . . . 114

9.4 Intention-to-treat effect: the effect of a misclassified treatment . 115

9.5 Per-protocol effect . . . . . . 117

10 Random variability 121

10.1 Identification versus estimation . . . . 121

10.2 Estimation of causal effects  . . 124

10.3 Themyth of the super-population . . . . . . 126

10.4 The conditionality “principle” . . . .  . . . 127

10.5 The curse of dimensionality . . . . . . . 131


II Causal inference with models 1

11 Why model? 3

11.1 Data cannot speak for themselves .. . . . 3

11.2 Parametric estimators .  . . . . . . . . . . 5

11.3 Nonparametric estimators . . . . . . . 6

11.4 Smoothing . . . .. . . . 7

11.5 The bias-variance trade-off . . .  . . . . 9

12 IP weighting and marginal structural models 11

12.1 The causal question . . . . . . . . . . . 11

12.2 Estimating IP weights viamodeling . . . . . 12

12.3 Stabilized IP weights . . . . . . .. . . . 14

12.4Marginal structuralmodels . . . . . 17

12.5 Effect modification andmarginal structural models . . . . 19

12.6 Censoring andmissing data. . . 20

13 Standardization and the parametric g-formula 23

13.1 Standardization as an alternative to IP weighting . . . 23

13.2 Estimating themean outcome viamodeling . . 25

13.3 Standardizing the mean outcome to the confounder distribution 26

13.4 IP weighting or standardization? . . . . 28

13.5 How seriously do we take our estimates? . . . . 29

14 G-estimation of structural nested models 31

14.1 The causal question revisited . .. . . 31

14.2 Exchangeability revisited . . .. . . . . . . . 32

14.3 Structural nestedmeanmodels . . . .. . . . . . 33

14.4 Rank preservation .  . . . . 35

14.5 G-estimation .. . . 37

14.6 Structural nestedmodels with two ormore parameters . . . . . 39

15 Outcome regression and propensity scores 43

15.1 Outcome regression . . . .. . . . 43

15.2 Propensity scores . . . . . . . . 45

15.3 Propensity stratification and standardization .  . . 46

15.4 Propensitymatching . . . . . . . . 48

15.5 Propensitymodels, structural models, predictivemodels . . . . 49

16 Instrumental variable estimation 53

16.1 The three instrumental conditions . . . .. . . 53

16.2 The usual IV estimand . . . . . . . . . . . . 56

16.3 A fourth identifying condition: homogeneity . .  . . 58

16.4 An alternative fourth condition: monotonicity . . . 61

16.5 The three instrumental conditions revisited . . .. . 63

16.6 Instrumental variable estimation versus other methods . . . . . 66

17 Causal survival analysis 69

17.1 Hazards and risks . . . . .. . 69

17.2 Fromhazards to risks . . .  . . 71

17.3Why censoringmatters . . . . . . . 74

17.4 IP weighting ofmarginal structural models. . . 76

17.5 The parametric g-formula  . . . . . 77

17.6 G-estimation of structural nestedmodels . . . 79

18 Variable selection for causal inference (coming in 2019) 83

III Causal inference from complex longitudinal data 1

19 Time-varying treatments 3

19.1 The causal effect of time-varying treatments . . . . 3

19.2 Treatment strategies .  . . . . . 4

19.3 Sequentially randomized experiments . . . . 5

19.4 Sequential exchangeability . . . . . . . . 8

19.5 Identifiability under some but not all treatment strategies . . . 9

19.6 Time-varying confounding and time-varying confounders . . . . 13

20 Treatment-confounder feedback 15

20.1 The elements of treatment-confounder feedback . . 15

20.2 The bias of traditional methods . . . . . . . 17

20.3Why traditionalmethods fail . .  . . . 19

20.4 Why traditional methods cannot be fixed . . . 21

20.5 Adjusting for past treatment . . . . . . 22

21 G-methods for time-varying treatments 25

21.1 The g-formula for time-varying treatments . . . . 25

21.2 IP weighting for time-varying treatments . .  . 28

21.3 A doubly robust estimator for time-varying treatments . . . . . 33

21.4 G-estimation for time-varying treatments . . . . 35

21.5 Censoring is a time-varying treatment .  . . 41



-END-

专 · 知

   专知开课啦!《深度学习: 算法到实战》, 中科院博士为你讲授!


请加专知小助手微信(扫一扫如下二维码添加),咨询《深度学习:算法到实战》参团限时优惠报名~

欢迎微信扫一扫加入专知人工智能知识星球群,获取专业知识教程视频资料和与专家交流咨询!

请PC登录www.zhuanzhi.ai或者点击阅读原文,注册登录专知,获取更多AI知识资料!

点击“阅读原文”,了解报名专知《深度学习:算法到实战》课程

登录查看更多
点赞 0
点赞 0
阅读2+
Top