Causal inference in statistics: An overview
By Judea Pearl. 2009
2,504 citations recorded by the source corpus.
This review presents empiricalresearcherswith recent advances in causal inference, and stresses the paradigmatic shifts that must be un- dertaken in moving from traditionalstatistical analysis to causal analysis of multivariate data. Special emphasis is placed on the assumptions that un- derly all causal inferences, the languages used in formulating those assump- tions, the conditional nature of all causal and counterfactual claims, and the methods that have been developed for the assessment of such claims. These advances are illustrated using a general theory of causation based on the Structural Causal Model (SCM) described in Pearl (2000a), which subsumes and unifies other approaches to causation, and provides a coher- ent mathematical foundation for the analysis of causes and counterfactuals. In particular, the paper surveys the development of mathematical tools for inferring (from a combination of data and assumptions) answers to three types of causal queries: (1) queries about the effects of potential interven- tions, (also called "causal effects" or "policy evaluation") (2) queries about probabilities of counterfactuals, (including assessment of "regret," "attri- bution" or "causes of effects") and (3) queries about direct and indirect effects (also known as "mediation"). Finally, the paper defines the formal and conceptual relationships between the structural and potential-outcome frameworks and presents tools for a symbiotic analysis that uses the strong features of both.
Subfields: Artificial Intelligence, Statistics and Probability, Causal Inference & Discovery
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Mohit Bansal
- Yoshua Bengio
- Percy Liang
- Bernhard Schölkopf
- Luke S. Zettlemoyer
- Jianfeng Gao
- Quoc V. Le
- Sergey Levine
In Statistics and Probability
- Bernhard Schölkopf
- Jan Peters
- Marloes H. Maathuis
- Peter Buhlmann
- Judea Pearl
- David Lopez-Paz
- Dominik Janzing
- Mihaela van der Schaar
In Causal Inference & Discovery
- Bernhard Schölkopf
- David Lopez-Paz
- Silvia Chiappa
- Jan Peters
- Xintao Wu
- Lu Zhang
- Dominik Janzing
- Judea Pearl
Citations in the corpus
- CausaLM: Causal Model Explanation Through Counterfactual Language Models (2020, methodology): “der to avoid the need in such a conditional generative model, we follow the bottom graph of Figure 3 and use an adversarial method, inspired by the idea of Process Control that was first introduced by Pearl (2009), to intervene on the text representation. We define the Textual Representation-based Ave”
- Counterfactual Fairness (2017, methodology): “In Pearl’s notation, where “ do ( A = a ) ” expresses an intervention on A at level a , we have that is a measure of causal effect, sometimes called the average causal effect (ACE).”
- Generative Counterfactual Introspection for Explainable Deep Learning (2019, background): “, they are neither discriminative nor counterfactual [7].”
- Counterfactual reasoning and learning systems: the example of computational advertising (2012, methodology): “Rubin (1986) argues that there is “no causation without manipulation”.”
- Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations (2020, background): “Note that while assumption 3 is standard across all methods for estimating treatment effects, it is not testable in practice (Robins et al., 2000; Pearl et al., 2009).”
- Counterfactual Memorization in Neural Language Models (2021, background): “Counterfactual is an important notion in statistical causality [Pearl et al., 2000, Rubin, 2005, Pearl, 2009, Imbens and Rubin, 2015] useful for studying causal probabilistic inference under alternative conditions.”
- Order-independent constraint-based causal structure learning (2012, background): “A DAG can be interpreted causally in the following way (Pearl, 2000, 2009; Spirtes et al., 2000):”
- The Bouncer Problem: Challenges to Remote Explainability (2019, methodology): “The classic way to assess a discrimination by the bouncer is for associations to run tests (following the principle of statistical causality [23] for instance): several persons attempt to enter, while they only vary in their attitude or appearance on the possibly discriminating feature ( e.g. , the…”
- A Survey on the Fairness of Recommender Systems (2022, methodology): “Counterfactual fairness requires that individuals have the same outcome in the real world as they do in the counterfactual world [72].”
- The future of FMRI connectivity (2012, background): “Experts in causal inference18 often draw a strong distinction between observational vs. interventional19 experimentation (for example, see the first half of (Pearl, 2009) by Judea Pearl, an accessible summary of concepts in causality).”
- Attachment disorganization and dissociation in virtual worlds: A study on problematic Internet use among players of online role playing games. (2012, methodology): “…pits specifies otherwise, the methodology used in this study and the evidence of mediation does not allow to logically rule out a reverse causation (Pearl 2009), i.e. it is also possible that the problematic internet use drives the attachment disorganization. furthermore, it may also be that the…”
- Estimating Causal Effects by Bounding Confounding (2014, background): “Back-door/front-door criterion (see [Pearl, 2000, 2009]): This approach applies for the case where some variables on the confounding path or between X and Y are measured and we know the causal structure underlying all variables together.”
- Exploring the Effects of COVID-19 Containment Policies on Crime: An Empirical Analysis of the Short-term Aftermath in Los Angeles (2020, background): “While Pearl (2009) demonstrates that post-facto observational studies cannot provide any answer to the investigation of true causal inference due to the potential presence of confounding factors, several quasi-experimental alternatives have been proposed to overcome the impossibility to run RCT in…”