## Transportability across studies: A formal approach (2010)

Citations: | 8 - 5 self |

### BibTeX

@TECHREPORT{Pearl10transportabilityacross,

author = {Judea Pearl and Elias Bareinboim},

title = {Transportability across studies: A formal approach},

institution = {},

year = {2010}

}

### OpenURL

### Abstract

We provide a formal definition of the notion of “transportability, ” or “external validity, ” which we view as a license to transfer causal information learned in experimental studies to a different environment, in which only observational studies can be conducted. We introduce a formal representation called “selection diagrams ” for expressing knowledge about differences and commonalities between populations of interest and, using this representation, we derive procedures for deciding whether causal effects in the target environment can be inferred from experimental findings in a different environment. When the answer is affirmative, the procedures identify the set of experimental and observational studies that need be conducted to license the transport. We further demonstrate how transportability analysis can guide the transfer of knowledge among non-experimental studies to minimize re-measurement cost and improve prediction power. We further provide a causally principled definition of “surrogate endpoint ” and show that the theory of transportability can assist the identification of valid surrogates in a complex network of cause-effect relationships. 1 Introduction: Threats

### Citations

551 |
Experimental and quasi-experimental designs for research
- DT, JC, et al.
- 1963
(Show Context)
Citation Context ... experiment is said to have “external validity” if the distribution of outcomes realized by a treatment group is the same as the distribution of outcome that would be realized in an actual program.” (=-=Campbell and Stanley, 1963-=-, p. 5) take a slightly broader view: “‘External validity’ asks the question of generalizability: To what population, settings, treatment variables, and measurement variables can this effect be genera... |

319 | Experimental and quasiexperimental designs for generalized casual inferences - Shadish, Cook, et al. - 2002 |

172 | Causal diagrams for empirical research
- Pearl
- 1995
(Show Context)
Citation Context ...rmal language within which the notion “environment” (or “population”) is given precise characterization, and differences among environments can be encoded and analyzed. The advent of causal diagrams (=-=Pearl, 1995-=-; Greenland et al., 1999; Spirtes et al., 2000; Pearl, 2009) provides such a language and renders the formalization of transportability possible. Armed with this language, this paper departs from the ... |

132 |
JM: Causal diagrams for epidemiologic research. Epidemiology
- Greenland, Pearl, et al.
- 1999
(Show Context)
Citation Context ... within which the notion “environment” (or “population”) is given precise characterization, and differences among environments can be encoded and analyzed. The advent of causal diagrams (Pearl, 1995; =-=Greenland et al., 1999-=-; Spirtes et al., 2000; Pearl, 2009) provides such a language and renders the formalization of transportability possible. Armed with this language, this paper departs from the tradition of communicati... |

74 | Direct and indirect effects - Pearl |

65 | Influence diagrams for causal modelling and inference - Dawid |

56 | A general identification condition for causal effects - Tian, Pearl |

52 |
Estimating mediated effects in prevention studies
- MacKinnon, Dwyer
- 1993
(Show Context)
Citation Context ...n other words, the conditions that license transport as well as 4 This is precisely the problem that motivated the unsettled literature on “surrogate endpoint” (Prentice, 1989; Freedman et al., 1992; =-=MacKinnon and Dwyer, 1993-=-; Fleming and DeMets, 1996; Burzykowski et al., 2005; Baker, 2006; MacKinnon et al., 2007; Joffe and Green, 2009), that is, finding a way of adjusting for a post-exposure variable Z so as to render ef... |

47 |
Principal stratification in causal inference
- Frangakis, Rubin
- 2002
(Show Context)
Citation Context ...1998); Buyse et al. (2000)) are based on regressions, they are unwittingly motivated by causal considerations, and several attempts have been made since to reformulate surrogacy in causal vocabulary (=-=Frangakis and Rubin, 2002-=-; Gilbert and Hudgens, 2008; Joffe and Green, 2009; Wolfson and Gilbert, 2010). At its core, the problem concerns a randomized trial where one seeks “endpoint surrogate,” namely, a variable that would... |

41 | Identification of conditional interventional distributions
- Shpitser, Pearl
- 2006
(Show Context)
Citation Context ...if part): If R is transportable, its transport formula T must satisfy the condition of Lemma 1, and that means that R = T is a valid equality in do-calculus. Moreover, since do-calculus is complete, (=-=Shpitser and Pearl, 2006-=-) every valid equality can be obtained by a finite application of the three rules of the calculus. This proves the Lemma. Definition 2 (Direct Transportability) A causal relation R is said to be direc... |

36 | Direct and indirect causal effects via potential outcomes - Rubin |

32 | Identification, inference and sensitivity analysis for causal mediation effects - Imai, Keele, et al. - 2010 |

32 |
Surrogate endpoints in clinical trials: Definitions and operational criteria
- Prentice
- 1989
(Show Context)
Citation Context ... is a causal, not statistical notion. In other words, the conditions that license transport as well as 4 This is precisely the problem that motivated the unsettled literature on “surrogate endpoint” (=-=Prentice, 1989-=-; Freedman et al., 1992; MacKinnon and Dwyer, 1993; Fleming and DeMets, 1996; Burzykowski et al., 2005; Baker, 2006; MacKinnon et al., 2007; Joffe and Green, 2009), that is, finding a way of adjusting... |

26 | Semantics of causal dag models and the identification of direct and indirect effects - Robins - 2003 |

24 | Identification for prediction and decision - Manski - 2007 |

22 | Estimation of direct causal effects - Petersen, Sinisi, et al. - 2006 |

16 |
The intermediate endpoint effect in logistic and probit regression. Clinical Trials
- MacKinnon, Lockwood, et al.
- 2007
(Show Context)
Citation Context ...lem that motivated the unsettled literature on “surrogate endpoint” (Prentice, 1989; Freedman et al., 1992; MacKinnon and Dwyer, 1993; Fleming and DeMets, 1996; Burzykowski et al., 2005; Baker, 2006; =-=MacKinnon et al., 2007-=-; Joffe and Green, 2009), that is, finding a way of adjusting for a post-exposure variable Z so as to render effect estimates transportable across populations with differing P(z|do(x)). A solution to ... |

15 |
Statistical validation of intermediate endpoints for chronic diseases
- Freedman, Graubard, et al.
- 1992
(Show Context)
Citation Context ...t statistical notion. In other words, the conditions that license transport as well as 4 This is precisely the problem that motivated the unsettled literature on “surrogate endpoint” (Prentice, 1989; =-=Freedman et al., 1992-=-; MacKinnon and Dwyer, 1993; Fleming and DeMets, 1996; Burzykowski et al., 2005; Baker, 2006; MacKinnon et al., 2007; Joffe and Green, 2009), that is, finding a way of adjusting for a post-exposure va... |

14 | The mediation formula: A guide to the assessment of causal pathways in nonlinear models - Pearl - 2012 |

13 | Geys H. The validation of surrogate endpoints in meta-analyses of randomized experiments - Buyse, Molenberghs, et al. |

13 |
Surrogate end points in clinical trials: are we being misled
- Fleming, DeMets
- 1996
(Show Context)
Citation Context ...ns that license transport as well as 4 This is precisely the problem that motivated the unsettled literature on “surrogate endpoint” (Prentice, 1989; Freedman et al., 1992; MacKinnon and Dwyer, 1993; =-=Fleming and DeMets, 1996-=-; Burzykowski et al., 2005; Baker, 2006; MacKinnon et al., 2007; Joffe and Green, 2009), that is, finding a way of adjusting for a post-exposure variable Z so as to render effect estimates transportab... |

12 | The validation of surrogate endpoints in randomized experiments - Buyse, Molenberghs - 1998 |

9 | Surrogate endpoints in clinical trials - Ellenberg, Hamilton - 1989 |

9 |
Related causal frameworks for surrogate outcomes
- Joffe, Greene
(Show Context)
Citation Context ...nsettled literature on “surrogate endpoint” (Prentice, 1989; Freedman et al., 1992; MacKinnon and Dwyer, 1993; Fleming and DeMets, 1996; Burzykowski et al., 2005; Baker, 2006; MacKinnon et al., 2007; =-=Joffe and Green, 2009-=-), that is, finding a way of adjusting for a post-exposure variable Z so as to render effect estimates transportable across populations with differing P(z|do(x)). A solution to this problem will be pr... |

6 | Estimating the proportion of treatment effect explained by a surrogate - Lin, Fleming, et al. |

6 |
Confounding equivalence in causal equivalence
- Pearl, Paz
- 2010
(Show Context)
Citation Context ...ains when we ask to transport the relation R ′ = ∑ P(y|x, z1)P(z1) (21) z1 Here we observe that, since Z1 and Z2 are each a sufficient set (i.e., back-door admissible), Z1 is interchangeable with Z2 (=-=Pearl and Paz, 2010-=-) and R ′ can be written as: R ′ = ∑ P(y|x, z2)P(z2) (22) z2 Therefore, using the independencies (S ⊥Y |X, Z2) and (S ⊥Z2) shown in the diagram, the transported relation becomes: R ′ (P ∗ ) = ∑ P ∗ (y... |

5 |
Generalizing evidence from randomized clinical trials to target populations
- Cole, Stuart
(Show Context)
Citation Context ...“standardization” – a statistical extrapolation method that can be traced back to a century-old tradition in demography and political arithmetic (Westergaard, 1916; Yule, 1934; Lane and Nelder, 1982; =-=Cole and Stuart, 2010-=-). On a second thought it raises the deeper question of why we consider age-specific effects to be invariant across populations. See discussion following Example 2. 3both, being a proxy for age which... |

5 | Analysis of covariance and standardization as instances of prediction. Biometrics - Lane, Nelder - 1982 |

5 | Discussion on causality - Lauritzen |

3 |
The Evaluation of Surrogate Endpoints
- Burzykowski, Molenberghs, et al.
- 2005
(Show Context)
Citation Context ...as well as 4 This is precisely the problem that motivated the unsettled literature on “surrogate endpoint” (Prentice, 1989; Freedman et al., 1992; MacKinnon and Dwyer, 1993; Fleming and DeMets, 1996; =-=Burzykowski et al., 2005-=-; Baker, 2006; MacKinnon et al., 2007; Joffe and Green, 2009), that is, finding a way of adjusting for a post-exposure variable Z so as to render effect estimates transportable across populations with... |

3 | Causal effects in psychotherapy: counterfactuals counteract overgeneralization. Psychother Res 2010 - Höfler, Gloster, et al. |

3 |
Scope and method of statistics
- Westergaard
- 1916
(Show Context)
Citation Context ..., Eq. (1) may be regarded as a routine application of “standardization” – a statistical extrapolation method that can be traced back to a century-old tradition in demography and political arithmetic (=-=Westergaard, 1916-=-; Yule, 1934; Lane and Nelder, 1982; Cole and Stuart, 2010). On a second thought it raises the deeper question of why we consider age-specific effects to be invariant across populations. See discussio... |

2 |
Surrogate endpoints: Wishful thinking or reality? National Cancer Institute 98 502–503
- Baker
- 2006
(Show Context)
Citation Context ...sely the problem that motivated the unsettled literature on “surrogate endpoint” (Prentice, 1989; Freedman et al., 1992; MacKinnon and Dwyer, 1993; Fleming and DeMets, 1996; Burzykowski et al., 2005; =-=Baker, 2006-=-; MacKinnon et al., 2007; Joffe and Green, 2009), that is, finding a way of adjusting for a post-exposure variable Z so as to render effect estimates transportable across populations with differing P(... |

2 |
Statistical identifiability and the surrogate endpoint problem with application to vaccine trials
- Wolfson, Gilbert
(Show Context)
Citation Context ...tivated by causal considerations, and several attempts have been made since to reformulate surrogacy in causal vocabulary (Frangakis and Rubin, 2002; Gilbert and Hudgens, 2008; Joffe and Green, 2009; =-=Wolfson and Gilbert, 2010-=-). At its core, the problem concerns a randomized trial where one seeks “endpoint surrogate,” namely, a variable that would allow good predictability of outcome for both treatment and control. In the ... |

1 |
Evaluating candidate principal surrogate endpoints
- Gilbert, Hudgens
- 2008
(Show Context)
Citation Context ... are based on regressions, they are unwittingly motivated by causal considerations, and several attempts have been made since to reformulate surrogacy in causal vocabulary (Frangakis and Rubin, 2002; =-=Gilbert and Hudgens, 2008-=-; Joffe and Green, 2009; Wolfson and Gilbert, 2010). At its core, the problem concerns a randomized trial where one seeks “endpoint surrogate,” namely, a variable that would allow good predictability ... |