Identifiability and Exchangeability for Direct and Indirect Effects

By James M. Robins, Sander Greenland. 1992

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Causal Inference & Discovery

Citations in the corpus

Causal Mediation Analysis for Interpreting Neural NLP: The Case of Gender Bias (2020, background): “Causal mediation analysis aims to measure how a treatment effect is mediated by intermediate variables (Robins and Greenland, 1992; Pearl, 2001; Robins, 2003).”
Fair Inference on Outcomes (2017, background): “Given Y ( a, M ( a ′ )) , we deﬁne the following effects on the mean difference scale: (Robins and Greenland 1992).”
Direct and Indirect Effects (2001, background): “Robins and Greenland (1992) have adapted essentially the same criterion (phrased differently) for their interpretation of “direct effect” in epidemiology.”
Causal inference in statistics: An overview (2009, background): “This hypothetical change, which Robins and Greenland (1992) called “pure” and Pearl (2001) called “natural,” mirrors what lawmakers instruct us to consider in race or sex discrimination cases: “The central question in any employment-discrimination case is whether the employer would have taken the…”
Theoretical Impediments to Machine Learning With Seven Sparks from the Causal Revolution (2018, neutral): “…problem, and its various manifestations are well recognized by machine-learning researchers, and enterprises such as “domain adaptation,” “transfer learning,” “life-long learning,” and “explainable AI,” are just some of the subtasks identiﬁed by researchers and funding agencies in an attempt to…”
A general approach to causal mediation analysis. (2010, unknown): “Following prior work (e.g., Robins and Greenland, 1992; Pearl, 2001; Glynn, 2008; Imai et al., 2008), we dene causal mediation eects using the potential outcomes notation.”
Mendelian randomisation for mediation analysis: current methods and challenges for implementation (2019, result): “If there is no interaction between the exposure and mediator, the estimate of the natu-ral direct effect is equivalent to the controlled direct effect, and indeed would align with estimates from Baron and Kenny approaches to mediation [4, 9, 16].”
Semiparametric Theory for Causal Mediation Analysis: efficiency bounds, multiple robustness, and sensitivity analysis (2012, background): “In words, the controlled direct effect refers to the exposure effect that arises upon intervening to set the mediator to a fixed level that may differ from its actual observed value [Robins and Greenland (1992), Pearl (2001), Robins (2003)].”
Causal mediation analysis for stochastic interventions (2019, methodology): “This is the same counterfactual variable that is often used to perform mediation analyses on the average treatment effect (Robins and Greenland, 1992; Pearl, 2001).”
Related Causal Frameworks for Surrogate Outcomes (2009, background): “(6) Direct and indirect effects in randomized trials may be estimated under two types of assumptions: initial randomization (Robins and Greenland, 1994; Ten Have et al., 2004; Joffe, Small, and Hsu, 2007) and sequential ignorability (Robins and Greenland, 1992).”
Interpretation and identification of causal mediation. (2014, unknown): “2 Natural Direct Effects: Identification As noted in (Robins and Greenland, 1992), we cannot generally evaluate the average natural direct effect from empirical data.”
On causal inference in the presence of interference (2010, background): “However once again, with (23), even if for each individual Aij were randomised conditional on Lij, h(Li), this does not guarantee that Yij(aij, v) is mean independent of v(ai, j) conditional on Lij, h(Li) unless h(Li)1⁄4Li.”
A Bayesian Semiparametric Approach to Intermediate Variables in Causal Inference (2011, methodology): “It is well documented that directly applying standard methods of pre-treatment variable adjustment, such as regression methods, to intermediate variables can result in estimates that generally lack causal interpretation (e.g., Rosenbaum 1984).”