## Causal Inference in the Health Sciences: A Conceptual Introduction (2001)

Venue: | Health Services and Outcomes Research Methodology |

Citations: | 8 - 0 self |

### BibTeX

@INPROCEEDINGS{Pearl01causalinference,

author = {Judea Pearl},

title = {Causal Inference in the Health Sciences: A Conceptual Introduction},

booktitle = {Health Services and Outcomes Research Methodology},

year = {2001},

pages = {2001},

publisher = {Kluwer Academic Publishers}

}

### OpenURL

### Abstract

This paper provides a conceptual introduction to causal inference, aimed to assist health services researchers benefit from recent advances in this area. The paper stresses the paradigmatic shifts that must be undertaken in moving from traditional statistical analysis to causal analysis of multivariate data. Special emphasis is placed on the assumptions that underlie all causal inferences, the languages used in formulating those assumptions, and the conditional nature of causal claims inferred from nonexperimental studies. These emphases are illustrated through a brief survey of recent results, including the control of confounding, corrections for noncompliance, and a symbiosis between counterfactual and graphical methods of analysis.

### Citations

7436 |
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference
- Pearl
- 1988
(Show Context)
Citation Context ...alth research community find it hard to appreciate and benefit from the many theoretical results that causal analysis has produced in the past two decades. These include advances in graphical models (=-=Pearl, 1988; Laurit-=-zen, 1996; Cowell et al., 1999), counterfactual or ‘‘potential outcome’’ analysis (Rosenbaum and Rubin, 1983; Robins, 1986; Manski, 1995; Angrist et al., 1996; Greenland et al., 1999b), struct... |

1458 |
The Central Role of the Propensity Score in Observational Studies for Causal Effects.” Biometrika 70(1
- Rosenbaum, Rubin
- 1983
(Show Context)
Citation Context ...causal analysis has produced in the past two decades. These include advances in graphical models (Pearl, 1988; Lauritzen, 1996; Cowell et al., 1999), counterfactual or ‘‘potential outcome’’ an=-=alysis (Rosenbaum and Rubin, 1983-=-; Robins, 1986; Manski, 1995; Angrist et al., 1996; Greenland et al., 1999b), structural equation models (Heckman ands190 PEARL Smith, 1998), and a more recent formulation, which unifies these approac... |

1244 |
Causality : models, reasoning, and inference
- Pearl
- 2000
(Show Context)
Citation Context ... and Sorbon, 1978), even though the causal content of these models has been obscured significantly since their inception (Muthen, 1987; Chou and Bentler, 1995) (see [Freedman, 1987] for critique and [=-=Pearl, 2000-=-, Chapter 5] for historical perspective). Section 3.2 uses these modeling fundamentals to develop simple mathematical tools for estimating causal effects and for the control of confounding. These tool... |

1206 |
Structural Equations with Latent Variables
- Bollen
- 1989
(Show Context)
Citation Context ...om Linear to Nonparametric Models Structural equation modeling (SEM) has been the main vehicle for effect analysis in economics and the behavioral and social sciences (Goldberger, 1972; Duncan, 1975; =-=Bollen, 1989-=-). However, the bulk of SEM methodology was developed for linear analysis and, until recently, no comparable methodology has been devised to extend its capabilities to models involving dichotomous var... |

1154 | Graphical Models
- Lauritzen
- 1996
(Show Context)
Citation Context ... community find it hard to appreciate and benefit from the many theoretical results that causal analysis has produced in the past two decades. These include advances in graphical models (Pearl, 1988; =-=Lauritzen, 1996; Cowell-=- et al., 1999), counterfactual or ‘‘potential outcome’’ analysis (Rosenbaum and Rubin, 1983; Robins, 1986; Manski, 1995; Angrist et al., 1996; Greenland et al., 1999b), structural equation mod... |

649 |
Estimating Causal Effects of Treatments in Randomized and Nonrandomized Studies
- Rubin
- 1974
(Show Context)
Citation Context ... can be captured in the powerful language of probability theory. How does one recognize causal expressions in the statistical literature? Those versed in the potential-outcome notation (Neyman, 1923; =-=Rubin, 1974; -=-Holland, 1988), can recognize such expressions through the subscripts that are attached to counterfactual events and counterfactual variables, e.g., Y xðuÞ or Z xy: (Some authors use parenthetical e... |

620 | Identification and Estimation of Local Average Treatment Effects - Imbens, Angrist - 1994 |

580 |
Identification of Causal Effects Using Instrumental Variables
- Angrist, Imbens, et al.
- 1996
(Show Context)
Citation Context ...hese include advances in graphical models (Pearl, 1988; Lauritzen, 1996; Cowell et al., 1999), counterfactual or ‘‘potential outcome’’ analysis (Rosenbaum and Rubin, 1983; Robins, 1986; Manski=-=, 1995; Angrist et al., 1996-=-; Greenland et al., 1999b), structural equation models (Heckman ands190 PEARL Smith, 1998), and a more recent formulation, which unifies these approaches under a single interpretation (Pearl, 1995a, 2... |

465 |
Graphical models in applied multivariate statistics
- Whittaker
- 1990
(Show Context)
Citation Context ... Note the b represents the incremental causal effect of X on Y, defined by b D EðYjdoðx 0 þ 1ÞÞ EðYjdoðx 0ÞÞ: Naturally, all attempts to give b statistical interpretation have ended in frust=-=ration (Whittaker, 1990;-=- Wermuth, 1992; Wermuth and Cox, 1993).s218 PEARL 20. Looser bounds were derived earlier by Robins (1989) and Manski (1990). 21. The inequality is sharp in the sense that every distribution Pðx; y; z... |

353 |
Nizam A. Applied regression analysis and other multivariable methods. Pacific Grove (CA): Brooks/Cole
- Kleinbaum, Kupper, et al.
(Show Context)
Citation Context ...t confounding bias cannot be detected or corrected by statistical methods alone, not even by the most sophisticated techniques that purport to ‘‘control for confounders,’’ such as stepwise sel=-=ection (Kleinbaum et al., 1998) or col-=-lapsibility-based methods (Grayson, 1987). One must make some assumptions regarding causal relationships in the problem, in particular about how the potential ‘‘confounders’’ affect other cova... |

215 | Conditional independence in Statistical Theory - Dawid - 1979 |

214 | A theory of inferred causation - Pearl, Verma - 1991 |

187 | Causal diagrams for empirical research
- Pearl
- 1995
(Show Context)
Citation Context ...st et al., 1996; Greenland et al., 1999b), structural equation models (Heckman ands190 PEARL Smith, 1998), and a more recent formulation, which unifies these approaches under a single interpretation (=-=Pearl, 1995-=-a, 2000). This paper aims at making these advances more accessible to the general research community. 1 To this end, Section 2 begins by illuminating two conceptual barriers that impede the transition... |

187 |
A new approach to causal inference in mortality studies with a sustained exposure period: Application to control of the healthy worker survivor effect
- Robins
- 1986
(Show Context)
Citation Context ...d in the past two decades. These include advances in graphical models (Pearl, 1988; Lauritzen, 1996; Cowell et al., 1999), counterfactual or ‘‘potential outcome’’ analysis (Rosenbaum and Rubin=-=, 1983; Robins, 1986-=-; Manski, 1995; Angrist et al., 1996; Greenland et al., 1999b), structural equation models (Heckman ands190 PEARL Smith, 1998), and a more recent formulation, which unifies these approaches under a si... |

167 |
Causal diagrams for epidemiologic research. Epidemiology
- Greenland, Pearl, et al.
- 1999
(Show Context)
Citation Context ...in graphical models (Pearl, 1988; Lauritzen, 1996; Cowell et al., 1999), counterfactual or ‘‘potential outcome’’ analysis (Rosenbaum and Rubin, 1983; Robins, 1986; Manski, 1995; Angrist et al.=-=, 1996; Greenland et al., 1999-=-b), structural equation models (Heckman ands190 PEARL Smith, 1998), and a more recent formulation, which unifies these approaches under a single interpretation (Pearl, 1995a, 2000). This paper aims at... |

166 | Correlation and causation - Wright - 1921 |

158 | Nonparametric Bounds on Treatment Effects - Manski - 1990 |

154 | Nature’s Capacities and Their Measurement - Cartwright - 1989 |

154 |
The Planning of Experiments
- Cox
- 1958
(Show Context)
Citation Context ...he verbal description with which investigators justify assumptions. For example, the assumption that a covariate is not affected by a treatment, a necessary assumption for the control of confounding (=-=Cox, 1958-=-), is expressed in plain English, not in a mathematical expression. The absence of notational distinction between causal and statistical relationships at first seemed harmless, because investigators w... |

132 |
A probabilistic theory of causality
- Suppes
- 1970
(Show Context)
Citation Context ...ent for determining which variables need be adjusted for (Pearl, 2000, pp. 182–189). 8. Attempts to define causal dependence by adding temporal information and conditioning on the entire past (e.g.,=-= [Suppes, 1970]) violate the-=- statistical requirement of limiting the analysis to ‘‘observed variables,’’ and encounter other insurmountable difficulties (see Eells [1991], Pearl [2000], pp. 249–257). 9. By ‘‘untest... |

125 | Identifiability and exchangeability for direct and indirect effects. Epidemiology - JM, Greenland - 1992 |

102 | Comment: Graphical models, causality and intervention
- Pearl
- 1993
(Show Context)
Citation Context ...rect treatment effect in that subpopulation of subjects. Such a set of factors is called a ‘‘sufficient set’’ or a set ‘‘appropriate for adjustment.’’ The following criterion, named ��=-=�‘back-door’’ in (Pearl, 1993-=-), provides a graphical method of selecting such a set of factors for adjustment. It states that a set S is appropriate for adjustment if two conditions hold: 1. No element of S is a descendant of X. ... |

87 |
Introduction to structural equation models
- Duncan
- 1975
(Show Context)
Citation Context ...ive that is relatively new to the health research literature. It is based on structural equation models (SEM), which have been used extensively in economics and the social sciences (Goldberger, 1972; =-=Duncan, 1975-=-; Joreskog and Sorbon, 1978), even though the causal content of these models has been obscured significantly since their inception (Muthen, 1987; Chou and Bentler, 1995) (see [Freedman, 1987] for crit... |

86 |
Causal inference, path analysis, and recursive structural equations models (with discussion
- Holland
- 1988
(Show Context)
Citation Context ...red in the powerful language of probability theory. How does one recognize causal expressions in the statistical literature? Those versed in the potential-outcome notation (Neyman, 1923; Rubin, 1974; =-=Holland, 1988),-=- can recognize such expressions through the subscripts that are attached to counterfactual events and counterfactual variables, e.g., Y xðuÞ or Z xy: (Some authors use parenthetical expressions, e.g... |

85 | The analysis of randomized and non-randomized aids treatment trials using a new approach to causal inference in longitudinal studies - Robins - 1989 |

79 | Causal ordering and identifiability - Simon - 1953 |

75 | Case-control studies. Design conduct analysis. 4th ed - Schlesselman - 1982 |

74 | Bounds on Treatment Effects From Studies with Imperfect Compliance
- Balke, Pearl
- 1997
(Show Context)
Citation Context ...ary to his or ð26ÞsCAUSAL INFERENCE IN THE HEALTH SCIENCES 211 her assignment. Under this assumption, which Imbens and Angrist (1994) call monotonicity, the inequalities in Eq. (26) can be tightened=-= (Balke and Pearl, 1997-=-) to give Pðy; X 1jZ 1Þ Pðy; X 1jZ 0Þ Pðy; X 0jZ 0Þ Pðy; X 0jZ 1Þ for all y 2f0; 1g: Violation of these inequalities now means either selection bias or a direct effect of Z on Y or the... |

71 | Instrumental Variables - Bowden, Turkington - 1984 |

69 | Probabilistic Causality - Eells - 1991 |

67 | Evaluating the welfare state. In - Heckman, Smith - 1998 |

63 | Probabilistic Evaluation of Sequential Plans from Causal Models with Hidden Variables.” Pp. 444–53 - Pearl, Robins - 1995 |

60 | Causal inference from graphical models - Lauritzen - 2001 |

59 | Recursive causal models - Kiiveri, Speed, et al. - 1984 |

58 | Theory of Probability: A Critical Introductory Treatment. Vol.1 - Finetti - 1970 |

55 | A graphical approach to the identification and estimation of causal parameters in mortality studies with sustained exposure periods - Robins - 1987 |

55 | Bayesian Inference for Causal Effects in Randomized Experiments With Noncompliance,” The Annals of Statistics - Imbens, Rubin - 1997 |

45 | Counterfactuals and policy analysis in structural models - Balke, Pearl - 1995 |

45 |
The Analysis of Case-Control Studies
- BRESLOW, DAY
- 1980
(Show Context)
Citation Context ...irst seemed harmless, because investigators were able to keep such distinctions implicitly in their heads, and managed to confine the mathematics to conventional, conditional probability expressions (=-=Breslow and Day, 1980; Mi-=-ettinen and Cook, 1981). However, as problem complexity grew, the notational inadequacy of probability calculus began to surface, and intense controversies ensued in the 1980–90’s between writers ... |

42 |
Linear dependencies represented by chain graphs
- Cox, Wermuth
- 1993
(Show Context)
Citation Context ...mental causal effect of X on Y, defined by b D EðYjdoðx 0 þ 1ÞÞ EðYjdoðx 0ÞÞ: Naturally, all attempts to give b statistical interpretation have ended in frustration (Whittaker, 1990; Wermuth=-=, 1992; Wermuth and Cox, 1993).-=-s218 PEARL 20. Looser bounds were derived earlier by Robins (1989) and Manski (1990). 21. The inequality is sharp in the sense that every distribution Pðx; y; zÞ satisfying Eq. (25) can be generated... |

38 |
Structural equation models in the social sciences
- Goldberger, Duncan
- 1973
(Show Context)
Citation Context ...ng from a perspective that is relatively new to the health research literature. It is based on structural equation models (SEM), which have been used extensively in economics and the social sciences (=-=Goldberger, 1972-=-; Duncan, 1975; Joreskog and Sorbon, 1978), even though the causal content of these models has been obscured significantly since their inception (Muthen, 1987; Chou and Bentler, 1995) (see [Freedman, ... |

37 | Testing identifiability of causal effects - Galles, Pearl - 1995 |

37 |
Confounding and collapsibility in causal inference
- Greenland, Pearl, et al.
- 1999
(Show Context)
Citation Context ...in graphical models (Pearl, 1988; Lauritzen, 1996; Cowell et al., 1999), counterfactual or ‘‘potential outcome’’ analysis (Rosenbaum and Rubin, 1983; Robins, 1986; Manski, 1995; Angrist et al.=-=, 1996; Greenland et al., 1999-=-b), structural equation models (Heckman ands190 PEARL Smith, 1998), and a more recent formulation, which unifies these approaches under a single interpretation (Pearl, 1995a, 2000). This paper aims at... |

36 | Identifiability, exchangeability and epidemiological confounding - Greenland, JM - 1986 |

36 | background knowledge in etiologic inference. Epidemiology 2001;11:313–20 - Data |

30 | Identifiability criteria for causal effects of joint interventions - Kuroki, Miyakawa - 1999 |

29 | Cause and counterfactual - Simon, Rescher - 1966 |

26 |
On identifying causal effects
- Tian, Shpitser
- 2010
(Show Context)
Citation Context ...ms involving multiples206 PEARL interventions (e.g., time varying treatments) were developed in Pearl and Robins (1995), Kuroki and Miyakawa (1999), and Pearl (2000, Chapters 3–4). A recent analysis=-= (Tian and Pearl, 2002-=-) further shows that the key to identifiability lies not in blocking paths between X and Y but, rather, in blocking paths between X and its immediate successors on the pathways to Y. All existing crit... |

26 | Conceptual problems in the definition and interpretation of attributable fractions - Greenland, JM - 1988 |

25 | The probability of causation under a stochastic model for individual risk - Robins, Greenland - 1989 |