Causal Inference Using Potential Outcomes

By Donald B. Rubin. 2005

1,689 citations recorded by the source corpus.

How researchers see this paper

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

In Statistics and Probability

Citations in the corpus

Causal inference by using invariant prediction: identification and confidence intervals (2015, background): “Among them are the framework of potential outcomes and counter-factuals [cf. Dawid, 2000, Rubin, 2005]; or structural equation modelling [cf. Bollen, 1989, Robins et al., 2000, Pearl, 2009] and corresponding graphical modeling [cf. Lauritzen and Spiegelhalter, 1988, Greenland et al., 1999, Spirtes…”
FLAME: A Fast Large-scale Almost Matching Exactly Approach to Causal Inference (2017, methodology): “We make Stable Unit Treatment Value (Rubin, 2005) and strong ignorability (Rosenbaum and Rubin, 1983) assumptions.”
Anchor regression: Heterogeneous data meet causality (2018, background): “In contrast, there is a growing literature on causal inference under various types of assumptions and different frameworks, with applications ranging from public health to biology and economics (Bollen, 1989; Dawid, 2000; Greenland et al., 1999; Lauritzen & Spiegelhalter, 1988; Pearl, 2009; Peters e”
The Mediation Myth (2015, methodology): “…for mediation analysis include inverse probability weighting (Robins, Hernán, & Brumback, 2000), the counterfactual or potential outcomes model (Rubin, 2005), principal stratification with propensity scores (Jo, Stuart, MacKinnon, & Vinokur, 2011), instrumental variable methods (Angrist,…”
Discovering Shifts to Suicidal Ideation from Mental Health Content in Social Media (2016, positive): “To do so, we work within the potential outcomes framework [57, 34] for causal modeling, applying stratified propensity score matching to estimate causal effects [56].”
The effects of quasi-random monetary experiments (2020, background): “In the parlance of the policy-evaluation literature (see, e.g., Rubin, 2005), any measure of the treatment effect of policy is contaminated by confounders simultaneously correlated with the treatment assignment mechanism and the outcome.”
Fair Decisions Despite Imperfect Predictions (2019, background): “In these settings, the decision typically determines which of the potential outcomes is observed and the focus is on confounders that effect both the decision and the outcome (Rubin, 2005).”
Handbook of Big Data (2016, methodology): “The software will be available as an R-package, but the instructions below do not reﬂect this and get the reader going by sourcing the software from the ﬁrst author’s site.”
Does Marriage Reduce Crime? (2015, background): “Central to this development is the increasing influence of the counterfactual model of causality, sometimes referred to as the “potential outcomes” framework (Rubin 2005), as an efficient analytical approach for specifying and estimating causal effects.”
Reward Tampering Problems and Solutions in Reinforcement Learning: A Causal Influence Diagram Perspective (2019, background): “Since the only paths from D2 and D3 to the utility nodes pass through the 23In Rubin-style potential outcomes notation (Rubin, 2005), our counterfactual variables D̃k correspond to (Dk)π=πsafe(S1,Θ R ∗ ), and the agent optimizes the sum of (Rk)D1:k=(D1:k)π=πsafe (S1,Θ R ∗ ) involving a double…”
A multidisciplinary survey on discrimination analysis (2013, background): “Propensity score analysis is also closely linked to causal analysis (Rubin, 2005), since it aims at answering the question ‘what kind of outcomes would we have observed had the stops involving protected drivers involved unprotected drivers instead?’.”
Fairness through Causal Awareness: Learning Causal Latent-Variable Models for Biased Data (2018, methodology): “y concerned with three endogenous variables in particular: X, the observable features (or covariates) of some example; T, a treatment which is applied to an example; and Y, the outcome of a treatment [38]. Our decision problem is: given an example with particular values for its features, X = x, what”
For objective causal inference, design trumps analysis (2008, background): “…observed outcome notation is inadequate in general, and can lead to serious errors—see, for example, the discussions in Holland and Rubin (1983) on Lord’s paradox, and in Rubin (2005), where errors are explicated that Fisher made because (I believe) he eschewed the potential outcome notation.”