Causal inference in statistics: An overview
Open accessThe DOI matches its Crossref record; 136 of the 136 references it cites were found in a registry or an index; 2 of the 72 sources its claims cite were checked for support.
Abstract
This review presents empirical researchers with recent advances in causal inference, and 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 underly all causal inferences, the languages used in formulating those assumptions, 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 coherent 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 interventions, (also called “causal effects” or “policy evaluation”) (2) queries about probabilities of counterfactuals, (including assessment of “regret,” “attribution” 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.
Overview
CoChat Summary
TL;DR
This survey argues that causal inference requires more than standard statistical association: it needs a formal language for causal assumptions and a mathematical machinery for combining those assumptions with data. Pearl presents the Structural Causal Model as a unifying framework that subsumes graphical, potential-outcome, and structural-equation approaches.
Key findings
- The paper distinguishes associational concepts (e.g., correlation, regression, conditional independence) from causal concepts (e.g., intervention, effect, confounding, attribution), arguing that causal claims cannot be defined from observed distributions alone.
- In Markovian structural models, the post-intervention distribution is identified by truncated factorization; for example, the paper shows that \(E(Y \mid do(x_0)) = E(Y \mid X=x_0)\) under the stated assumptions.
- The back-door criterion provides a graphical rule for choosing admissible adjustment sets; the paper states that if a set \(S\) blocks all back-door paths from \(X\) to \(Y\), then stratifying on \(S\) removes confounding bias.
- For noncompliance in clinical trials, the paper says treatment effects are generally nonidentifiable, but bounds can be derived; Balke and Pearl (1997) obtained closed-form bounds for 16 equivalent classes in the binary case, and the paper notes that these bounds can yield “significant and sometimes accurate” information.
Methods
- This is a conceptual review/survey, not an empirical study; it synthesizes advances in causal analysis “from the unifying perspective” of the Structural Causal Model.
- It develops the formalism using structural equations, causal diagrams, do-operator interventions, and counterfactual notation, and then maps these ideas to the potential-outcome framework.
- The analysis is mainly theoretical and illustrative, using examples such as mediation, instrumental variables, covariate adjustment/back-door identification, and noncompliance in clinical trials.
Limitations
- The paper states that counterfactual probabilities conditioned on outcomes “can in general [not] be identified from observational or even experimental studies” without additional conditions.
- It states that natural direct effects “cannot in general be identified or estimated, even with the help of ideal, controlled experiments” unless extra assumptions hold.
- Evident from the text: this is a survey and framework paper, so it does not report a new dataset, sample size, or empirical effect estimate.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
2 Valid · 1 Not valid · 53 Not checked · 133 Own results and interpretations
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Note also that the truncated factorization formula is not restricted to interventions on a single variable; it is applicable to simultaneous or sequential interventions such as those invoked in the analysis of time varying treatment with time varying confounders (Robins, 1986; Arjas and Parner, 2004).”
What the sources say
- Supportedcites Robins, 1986 source
Section F describes treatments assigned at successive times, with assignment depending on past measured covariates. The G-Computation Algorithm multiplies conditional probabilities along treatment and covariate-history paths and sums across those paths to calculate outcomes under a hypothetical treatment. This supports the cited point that the calculation extends beyond an intervention on one variable, although the paper calls it the G-computation algorithm rather than the truncated factorization formula. (LLM, conf=0.87)
- Not checked — no full text of the sourcecites Arjas and Parner, 2004 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The selection criterion can be applied systematically to diagrams of any size and shape, thus freeing analysts from judging whether ' X is conditionally ignorable given S ,' a formidable mental task required in the potential-response framework (Rosenbaum and Rubin, 1983).”
What the sources say
- Supportedcites Rosenbaum and Rubin, 1983 source
The citation supports the sentence’s reference to conditional ignorability in the potential-response framework. Section 1-2 defines strongly ignorable treatment assignment using the independence of treatment assignment from potential responses given covariates. The diagrammatic selection criterion is the citing authors’ own contrast, not a method attributed to Rosenbaum and Rubin. (LLM, conf=0.88)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“A few but challenging algebraic steps would reveal that one can perform the summation over Z 2 to obtain”
- Also stated in: § 3.3. Coping with unmeasured confounders · plain
“which means that we need only adjust for Z 1 and Z 3 without ever measuring Z 2 .”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 3.3. Coping with unmeasured confounders · plain
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The set { Z 3 } , however, is not sufficient for adjustment because, as explained above, it does not block the path X ← W 1 ← Z 1 → Z 3 ← Z 2 → W 2 → Y .”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“This expression coincides with Robins' (1987) G -computation formula, which was derived from a more complicated set of (counterfactual) assumptions.”
What the sources say
- Not checked — no full text of the sourcecites Robins, 1987 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
114 more claims
Create a free account to see all 189 claims and to verify the 7 references still open.
- Probably not a substantial claim. 39
- It may state two separate claims as one. 39
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“which is known as the truncated factorization formula , see for example, Corollary 1 in Pearl (2009).”
From Discrimination-Free Insurance Pricing (2022) · cites this paper as Pearl (2009)
In the citing paper
““observational data” (such as rfMRI) is in general not a robust and safe tool for inferring causality, compared with “interventional” studies (such as task-fMRI experiments) [63].”
From Functional connectomics from resting-state fMRI (2013) · cites this paper as [63]
Citations
Cites
The central role of the propensity score in observational studies for causal effects
ROSENBAUM, P. R., & RUBIN, D. B. (1983) · Biometrika · doi:10.1093/biomet/70.1.41 · Cited by 31,241 · cited 3×
“The selection criterion can be applied systematically to diagrams of any size and shape, thus freeing analysts from judging whether “X is conditionally ignorable given S,” a formidable mental task required in the potential-response framework (Rosenbaum and Rubin, 1983).”
BackgroundMethod(2001) · RePEc: Research Papers in Economics · Cited by 27,577
Structural Equations with Latent Variables.
Clogg, C. C., & Bollen, K. A. (1991) · Contemporary Sociology · doi:10.2307/2072165 · Cited by 17,442
Structural Equations with Latent Variables
Bollen, K. A. (1989) · Wiley series in probability and statistics · doi:10.1002/9781118619179 · Cited by 17,277 · cited 3×
“…identification has received considerable attention in econometrics (Hurwicz, 1950; Marschak, 1950; Koopmans, 1953) and social science (Duncan, 1975; Bollen, 1989), usually in linear parametric settings, were it reduces to asking whether some model parameter, β, has a unique solution in terms of…”
Causality: Models, Reasoning, and Inference
Pearl, J. (2009) · doi:10.1017/cbo9780511803161 · Cited by 11,653
Self-citationMediation in experimental and nonexperimental studies: New procedures and recommendations.
Shrout, P. E., & Bolger, N. (2002) · Psychological Methods · doi:10.1037/1082-989x.7.4.422 · Cited by 11,335 · cited 2×
“…) = ∑ z E(Y |x, z)[P (z|x′)− P (z|x)] (52) This expression deserves the label Mediation Formula, due to its pivotal role in mediation analysis (Imai et al., 2008), which has been a thorny issue in several sciences (Shrout and Bolger, 2002; MacKinnon et al., 2007; Mortensen et al., 2009).”
BackgroundEstimating causal effects of treatments in randomized and nonrandomized studies
Rubin, D. B. (1974) · Journal of Educational Psychology · doi:10.1037/h0037350 · Cited by 9,436 · cited 4×
“…equation models (SEM) used in economics and social science (Goldberger, 1973; Duncan, 1975), the potentialoutcome framework of Neyman (1923) and Rubin (1974), and the graphical models developed for probabilistic reasoning and causal analysis (Pearl, 1988; Lauritzen, 1996; Spirtes et al., 2000;…”
BackgroundMethodInfluentialMacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007) · Annual Review of Psychology · doi:10.1146/annurev.psych.58.110405.085542 · Cited by 5,688 · cited 2×
“…) = ∑ z E(Y |x, z)[P (z|x′)− P (z|x)] (52) This expression deserves the label Mediation Formula, due to its pivotal role in mediation analysis (Imai et al., 2008), which has been a thorny issue in several sciences (Shrout and Bolger, 2002; MacKinnon et al., 2007; Mortensen et al., 2009).”
BackgroundDiscrete Multivariate Analysis: Theory and Practice.
Beniger, J. R., Bishop, Y. M. M., Feinberg, S. E., Holland, P. W., Light, R. J., & Mosteller, F. (1975) · Contemporary Sociology · doi:10.2307/2063625 · Cited by 5,108
Recent Developments in the Econometrics of Program Evaluation
Imbens, G. W., & Wooldridge, J. M. (2009) · Journal of Economic Literature · doi:10.1257/jel.47.1.5 · Cited by 4,926
Research works printed but not linked: 11
- Glymour, M. and Greenland, S. (2008). Causal diagrams. In, Modern Epidemiology (K. Rothman, S. Greenland and T. Lash, eds.), 3rd ed. Lippincott Williams & Wilkins, Philadelphia, PA, 183–209.
- Hurwicz, L. (1950). Generalization of the concept of identification. In, Statistical Inference in Dynamic Economic Models (T. Koopmans, ed.). Cowles Commission, Monograph 10, Wiley, New York, 245–257.
- Koopmans, T. (1953). Identification problems in econometric model construction. In, Studies in Econometric Method (W. Hood and T. Koopmans, eds.). Wiley, New York, 27–48.
- Marschak, J. (1950). Statistical inference in economics. In, Statistical Inference in Dynamic Economic Models (T. Koopmans, ed.). Wiley, New York, 1–50. Cowles Commission for Research in Economics, Monograph 10.
- Pearl, J. (1988)., Probabilistic Reasoning in Intelligent Systems. Morgan Kaufmann, San Mateo, CA.
- Pearl, J. (2001). Direct and indirect effects. In, Proceedings of the Seventeenth Conference on Uncertainty in Artificial Intelligence. Morgan Kaufmann, San Francisco, CA, 411–420.
- Robins, J. (1989). The analysis of randomized and non-randomized aids treatment trials using a new approach to causal inference in longitudinal studies. In, Health Service Research Methodology: A Focus on AIDS (L. Sechrest, H. Freeman and A. Mulley, eds.). NCHSR, U.S. Public Health Service, Washington, D.C., 113–159.
- Wermuth, N. (1992). On block-recursive regression equations., Brazilian Journal of Probability and Statistics (with discussion) 6 1–56.
- Cole, P. (1997). Causality in epidemiology, health policy, and law., Journal of Marketing Research 27 10279–10285.
- Pearson, K., Lee, A. and Bramley-Moore, L. (1899). Genetic (reproductive) selection: Inheritance of fertility in man., Philosophical Transactions of the Royal Society A 73 534–539.
- Robertson, D. (1997). The common sense of cause in fact., Texas Law Review 75 1765–1800.
Cited by
Items 401–500 of 2,354
Semi-mechanistic Bayesian modelling of COVID-19 with renewal processes
Bhatt, S., Ferguson, N. M., Flaxman, S. R., Gandy, A., Mishra, S., & Scott, J. A. (2023) · Journal of the Royal Statistical Society Series A (Statistics in Society) · doi:10.1093/jrsssa/qnad030 · Cited by 69
Implicit Realism Impedes Progress in Psychology: Comment on Fried (2020)
Yarkoni, T. (2020) · Psychological Inquiry · doi:10.1080/1047840x.2020.1853478 · Cited by 32
Chen, S., Bi, K., Lyu, S., Sun, P., & Bonanno, G. A. (2022) · European Journal of Psychotraumatology · doi:10.1080/20008066.2022.2115635 · Cited by 37
“As the DAG is directed and has no feedback loop (i.e. acyclic), it provides preliminary inference about the causal relationships among nodes (Pearl, 2009).”
BackgroundThe algorithmization of counterfactuals
Pearl, J. (2011) · Annals of Mathematics and Artificial Intelligence · doi:10.1007/s10472-011-9247-9 · Cited by 46
“These tasks are performed using a general estimator, called the Mediation Formula [18, 19, 21], which is applicable to nonlinear models with discrete or continuous variables, and permits the evaluation of path-specific effects with minimal assumptions regarding the data-generating process.”
MethodSelf-citationGalas, D. J., Sakhanenko, N. A., Skupin, A., et al. (2013) · Journal of Computational Biology · doi:10.1089/cmb.2013.0039 · Cited by 30
“The introduction of directionality in the hypergraphs could be used to extend the present theory into the realm of representation of causality models as discussed by Pearl (2009, 2010).”
Bica, I., Alaa, A. M., Jordon, J., & van der Schaar, M. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.04083 · Cited by 282
“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).”
BackgroundNaser, M. Z. (2024) · Wiley Interdisciplinary Reviews Data Mining and Knowledge Discovery · doi:10.1002/widm.1533 · Cited by 37
“The d-separation criterion is a close companion to the Markovian assumption by establishing whether X is independent of Y , given Z (i.e., X ⫫ Y j Z ), by associating independence with the separation of variables in a causal graph (Pearl, 2009).”
Ness, P. H. V., Fried, T. R., & Gill, T. M. (2011) · Journal of Mixed Methods Research · doi:10.1177/1558689811412973 · Cited by 47
“Practitioners of qualitative data analysis in the biomedical field have been fairly cautious about proposing causal theories, and in fact, new causal quantitative methods have yet to find wide application in biomedical research (Pearl, 2009); social scientists have been more ambitious in this regard and have developed…”
BackgroundTowards integrative causal analysis of heterogeneous data sets and studies
Tsamardinos, I., Triantafillou, S., & Lagani, V. (2012) · Journal of Machine Learning Research · Cited by 43
Nielsen, D. V., Torp-Pedersen, C. T., Skals, R. K., Gerds, T. A., Karaliunaite, Z., & Jakobsen, C.-J. (2018) · Critical Care · doi:10.1186/s13054-018-1969-1 · Cited by 29
Quantifying Causal Influences in the Presence of a Quantum Common Cause
Gachechiladze, M., Miklin, N., & Chaves, R. (2020) · Physical Review Letters · doi:10.1103/physrevlett.125.230401 · Cited by 45
“[3] Judea Pearl et al.”
BackgroundNorbury, A., Liu, S. H., Campaña-Montes, J. J., Romero-Medrano, L., Barrigón, M. L., Smith, E. C., MEmind Study Group, Aroca, F., Artés-Rodríguez, A., Baca‐García, E., et al. (2020) · Molecular Psychiatry · doi:10.1038/s41380-020-00963-5 · Cited by 35
Priority Challenges for Social and Behavioral Research and Its Modeling
Davis, P. K., O’Mahony, A., Gulden, T. R., Osoba, O., & Sieck, K. (2018) · RAND Corporation eBooks · doi:10.7249/rr2208 · Cited by 28
Model selection in occupancy models: Inference versus prediction
Stewart, P. S., Stephens, P. A., Hill, R. A., Whittingham, M. J., & Dawson, W. (2022) · Ecology · doi:10.1002/ecy.3942 · Cited by 39
Robust Emotion Recognition in Context Debiasing
Yang, D., Yang, K., Li, M., Wang, S., Wang, S., & Zhang, L. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.01183 · Cited by 59
“Intervention [36] aims to actively manipulate the probability distributions of variables to obtain unbiased estimations or discover confounder effects.”
BackgroundNeugebauer, R. S., Schmittdiel, J. A., & van der Laan, M. J. (2016) · The International Journal of Biostatistics · doi:10.1515/ijb-2015-0028 · Cited by 33
“If the MSM framework above (missing data framework) is not explicitly resting on the more general structural framework through additional explicit assumptions encoded by a causal diagram [50], then an additional assumption referred to as consistency assumption is made [51, 52].”
Li, S., Vlassis, N. A., Kawale, J., & Fu, Y. (2016) · International Joint Conference on Artificial Intelligence · Cited by 35
Ferrari, E., Bosco, P., Calderoni, S., Oliva, P., Palumbo, L., Spera, G., Fantacci, M. E., & Retico, A. (2020) · Artificial Intelligence in Medicine · doi:10.1016/j.artmed.2020.101926 · Cited by 34
The Deductive Approach to Causal Inference
Pearl, J. (2014) · Journal of Causal Inference · doi:10.1515/jci-2014-0016 · Cited by 28
“Such a set of factors is called a “sufficient set,” “admissible set,” or a set “appropriate for adjustment” [9–11].”
BackgroundInfluentialSelf-citationNot All Clinical AI Monitoring Systems Are Created Equal: Review and Recommendations
Feng, J., Xia, F., Singh, K., & Pirracchio, R. (2025) · NEJM AI · doi:10.1056/aira2400657 · Cited by 28
Causality for Trustworthy Artificial Intelligence: Status, Challenges and Perspectives
Rawal, A., Raglin, A. J., Rawat, D. B., Sadler, B. M., & McCoy, J. (2025) · ACM Computing Surveys · doi:10.1145/3665494 · Cited by 58
“In structural causal models, causal efect of any variable can be investigated, making them the preferred framework for learning causal relations between variables [3].”
InfluentialEdelman, S. (2015) · Journal of Experimental & Theoretical Artificial Intelligence · doi:10.1080/0952813x.2015.1042534 · Cited by 34
“Causal learning is particularly important for goal-directed as opposed to habitual control, because representations that make causal dependencies explicit afford informed intervention (Chater & Oaksford, 2013; Holyoak & Cheng, 2011; Pearl, 2009).”
BackgroundDatabase Repair Meets Algorithmic Fairness
Salimi, B., Howe, B., & Suciu, D. M. (2020) · ACM SIGMOD Record · doi:10.1145/3422648.3422657 · Cited by 36
“However, it is known that individual-level counterfactuals can not be estimated from data in general [26].”
BackgroundMethodBarry, C.-J. S., Liu, J., Richmond, R. C., Rutter, M. K., Lawlor, D. A., Dudbridge, F., & Bowden, J. (2021) · PLoS Genetics · doi:10.1371/journal.pgen.1009703 · Cited by 31
“Since X is a function of both G and U, conditioning on X induces a correlation between them [35].”
BackgroundBITES: balanced individual treatment effect for survival data
Schrod, S., Schäfer, A., Solbrig, S., Lohmayer, R., Gronwald, W., Oefner, P. J., Beißbarth, T., Spang, R., Zacharias, H. U., & Altenbuchinger, M. (2022) · Bioinformatics · doi:10.1093/bioinformatics/btac221 · Cited by 40
“T j X : (4) This assumption ensures that the causal effect is not influenced by non-observable causal substructures such as confounding (Pearl, 2009).”
System Identification in Dynamic Networks
Dankers, A. G. (2014) · Research Repository (Delft University of Technology) · doi:10.4233/uuid:cb7f49da-8c2e-4370-bc6b-80af0684e372 · Cited by 45
“In statistics, and in particular in statistical inference, a confounding variable is a variable that is not known (or measured) and causally affects both the output variable and the input variable (Pearl, 2009).”
BackgroundWhen is invariance useful in an Out-of-Distribution Generalization problem ?Preprint
Koyama, M., & Yamaguchi, S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2008.01883 · Cited by 122
De Choudhury, M., & Kiciman, E. (2018) · AI Magazine · doi:10.1609/aimag.v39i3.2815 · Cited by 27
“In prior joint work (De Choudhury et al. 2016), the authors developed a causal inference framework (Pearl 2009) to assess the likelihood that an individual will transition to discussions of suicidal ideation, given a history of mental health discourse on social media.”
BackgroundGenetic Susceptibility to Gestational Diabetes Mellitus in a Chinese Population
Cao, M., Zhang, L., Chen, T., Shi, A., Xie, K., Li, Z., Xu, J., Chen, Z., Ji, C., & Wen, J. (2020) · Frontiers in Endocrinology · doi:10.3389/fendo.2020.00247 · Cited by 31
Skill-based Career Path Modeling and Recommendation
Ghosh, A., Woolf, B. P., Zilberstein, S., & Lan, A. (2020) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata50022.2020.9377992 · Cited by 34
“For all datasets, we sort user experiences based on their start times in chronological order.”
MethodCausal Support: Modeling Causal Inferences with Visualizations
Kale, A. M., Wu, Y., & Hullman, J. R. (2021) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2021.3114824 · Cited by 32
“DAGs encode hypothesized relationships among variables (e.g., Fig.1 B (cid:13) ), making causal relationships and the assumptions they entail explicit and in some cases testable [3, 41–43].”
Alipanahi, B., & Frey, B. J. (2013) · Nature Biotechnology · doi:10.1038/nbt.2657 · Cited by 29
Fecal Metabolites as Biomarkers for Predicting Food Intake by Healthy Adults
Shinn, L. M., Mansharamani, A., Baer, D. J., Novotny, J. A., Charron, C. S., Khan, N. A., Zhu, R., & Holscher, H. D. (2022) · Journal of Nutrition · doi:10.1093/jn/nxac195 · Cited by 32
“In fact, parallel-arm studies can prevent carryover effects between conditions that can affect causal inferences (61, 62).”
Vascularized organ bioprinting: From strategy to paradigm
Wang, H., Liu, X., Gu, Q., & Zheng, X. (2023) · Cell Proliferation · doi:10.1111/cpr.13453 · Cited by 29
Cankaya, B., Topuz, K., Delen, D., & Glassman, A. (2023) · Omega · doi:10.1016/j.omega.2023.102906 · Cited by 35
“In prediction problems needing causal inference, Bayesian Networks (BNs) are advantageous to other ML techniques by simply explaining complex inferences [47] .”
Gıovanıs, E., & Özdamar, Ö. (2016) · Journal of Economic Surveys · doi:10.1111/joes.12163 · Cited by 33
Robust Causal Structure Learning with Some Hidden Variables
Frot, B., Nandy, P., & Maathuis, M. H. (2019) · Journal of the Royal Statistical Society Series B (Statistical Methodology) · doi:10.1111/rssb.12315 · Cited by 60
DeYoung, C. G., & Blain, S. D. (2020) · Cambridge University Press eBooks · doi:10.1017/9781108264822.026 · Cited by 38
Explaining black box decisions by Shapley cohort refinementPreprint
Mase, M., Owen, A. B., & Seiler, B. B. (2019) · RePEc: Research Papers in Economics · doi:10.48550/arxiv.1911.00467 · Cited by 66
“Importance by one of these criteria need not imply importance by another, though additional assumptions may allow a causal implication to be made from one of the other measures (Pearl, 2009; Zhao and Hastie, 2019).”
BackgroundWilliams, D. R. (2018) · doi:10.31234/osf.io/x8dpr · Cited by 27
Williams, D. R., & Mulder, J. (2019) · doi:10.31234/osf.io/ypxd8 · Cited by 27
Limited-Data SAR ATR Causal Method via Dual-Invariance Intervention
Wang, C., Xu, R., Huang, Y., Pei, J., Huang, C., Zhu, W., & Yang, J. (2025) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2025.3528464 · Cited by 45
Endo, W., Santos, F. P., Simpson, D., Maciel, C. D., & Newland, P. L. (2015) · Journal of Computational Neuroscience · doi:10.1007/s10827-015-0548-6 · Cited by 26
Learning Opinion Dynamics From Social Traces
Monti, C., De Francisci Morales, G., & Bonchi, F. (2020) · Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3394486.3403119 · Cited by 31
“, regression), for which a causal interpretation is much harder [22].”
BackgroundMore, S. J., Bambidis, V., Benford, D., Bragard, C., Hernández, A. F., Bennekou, S. H., Koutsoumanis, K., Machera, K. A., Naegeli, H., Nielsen, S. S., et al. (2020) · EFSA Journal · doi:10.2903/j.efsa.2020.6221 · Cited by 27
“The subject of causality has also been elaborated by Pearl (2009) and Pearl and Mackenzie (2018).”
BackgroundDeconfounded recommendation via causal intervention
Yu, D., Li, Q., Wang, X., & Xu, G. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.01.089 · Cited by 31
Decoupled Unbiased Teacher for Source-Free Domain Adaptive Medical Object Detection
Liu, X., Li, W., & Yuan, Y. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3272389 · Cited by 40
“To alleviate the undesired phenomenon and achieve unbiased SFDA medical object detection, we refer to the causal mechanism in debiasing [22], [23], [24], [25] and frame an SCM for this task as illustrated in Fig.”
BackgroundMethodInfluentialOn Discrimination Discovery Using Causal Networks
Zhang, L., Wu, Y., & Wu, X. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-39931-7_9 · Cited by 29
“As defined in [15], a node set S not containing C or E blocks a path p between nodes C and E if either (1) p contains at least one noncollider X in set S, or (2) p contains at least one collider X, and X and all its descendants are outside set S.”
BackgroundMethodRobust Cyberbullying Detection with Causal Interpretation
Cheng, L., Guo, R., & Liu, H. (2019) · International World Wide Web Conference (WWW) · doi:10.1145/3308560.3316503 · Cited by 29
“Confounding bias can lead to Simpson’s paradox, a phenomenon whereby the association coefficient be-tween a covariate and the outcome reverses sign upon conditioning on a second covariate, regardless of the value of the second co-variate [20].”
BackgroundOmitted Variable Bias in GLMs of Neural Spiking Activity
Stevenson, I. H. (2018) · Neural Computation · doi:10.1162/neco_a_01138 · Cited by 25
“At the same time, although it has long been a concern in statistics (Pearl, 2009) and has received some attention in other fields (Clarke, 2005; Greenland, 1989), omitted variable bias, as a general problem, appears underappreciated in systems neuroscience.”
BackgroundShang, Z., Liang, D., Niu, L., et al. (2019) · Energy & Fuels · doi:10.1021/acs.energyfuels.9b01405 · Cited by 26
Causal Discovery for Relational Domains: Representation, Reasoning, and Learning
Maier, M. E. (2014) · Scholarworks (University of Massachusetts Amherst) · doi:10.7275/6041512.0 · Cited by 29
Soleimani, H., Subbaswamy, A., & Saria, S. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1704.02038 · Cited by 67
“Typically, the effect can be quantified as a difference in potential outcomes: the outcome under treatment and what the outcome would have been if no treatment were given (Rubin, 1974; Pearl et al., 2009).”
BackgroundReusing Natural ExperimentsPreprint
Heath, D., Ringgenberg, M. C., Samadi, M., & Werner, I. M. (2019) · SSRN Electronic Journal · doi:10.2139/ssrn.3457525 · Cited by 32
Counterfactual Data Augmentation using Locally Factored Dynamics
Pitis, S., Creager, E., & Garg, A. (2020) · arXiv · doi:10.48550/arxiv.2007.02863 · Cited by 26
“To answer the counterfactual question, “what might the transition at time t have looked like if component S t had value x instead of value y?”, we would ordinarily apply Pearl’s do-calculus to M to obtain submodelMdo(Si t=x) = 〈V,U,Fx〉, where Fx = F \ f i ∪ {S t = x} and incoming edges to S t are removed from Gdo(Si t…”
MethodBoelaert, J., & Ollion, É. (2018) · Revue Française de Sociologie · doi:10.3917/rfs.593.0475 · Cited by 25
Physics-Guided Deep Learning for Dynamical Systems: A Survey
Wang, R., & Yu, R. (2025) · ACM Computing Surveys · doi:10.1145/3766887 · Cited by 167
Clarifying the causes of consistent and inconsistent findings in genetics
Dattani, S., Howard, D. M., Lewis, C. M., & Sham, P. C. (2022) · Genetic Epidemiology · doi:10.1002/gepi.22459 · Cited by 29
“In this paper, we elucidate these concerns using structural causal models (Pearl, 2009).”
MethodSun, R., Schiavon, S., Brager, G. S., Arens, E. A., Zhang, H., Parkinson, T., & Zhang, C. (2024) · Building and Environment · doi:10.1016/j.buildenv.2024.111530 · Cited by 27
Improving disaster resilience with causal machine learning for flood damage estimation
Museru, M. L., Nazari, R., Nikoo, M. R., & Karimi, M. (2025) · The Science of The Total Environment · doi:10.1016/j.scitotenv.2025.180121 · Cited by 29
Harnessing data science to advance radiation oncology
Vogelius, I. R., Petersen, J., & Bentzen, S. M. (2020) · Molecular Oncology · doi:10.1002/1878-0261.12685 · Cited by 29
“Here, the most robust method is the randomized controlled trial, but other methods exist where randomization is not feasible (Pearl, 2009).”
MethodHosseini, M. M., Zargoush, M., Alemi, F., & Kheirbek, R. (2020) · Journal Of Big Data · doi:10.1186/s40537-020-00302-z · Cited by 29
Direct and indirect influences of political regimes on corruption
Goel, R. K., & Nelson, M. A. (2021) · Social Science Quarterly · doi:10.1111/ssqu.13006 · Cited by 28
Jena, P. R., & Grote, U. (2022) · Frontiers in Sustainable Food Systems · doi:10.3389/fsufs.2021.716904 · Cited by 25
“Several matching methods are proposed in the literature such as k nearest neighbor matching, radius matching, kernel matching, and Mahalanobis matching (Rosenbaum and Rubin, 1983; Heckman et al., 1998; Dehejia and Wahba, 2002; Pearl, 2009).”
MethodTschantz, M. C. (2022) · 2022 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3531146.3533242 · Cited by 25
Learning Distortion Invariant Representation for Image Restoration from a Causality Perspective
Li, X., Li, B., Jin, X., Lan, C., & Chen, Z. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.00171 · Cited by 46
“However, a spurious correlation [46] is also captured which introduces the bad confounding effects of specific degradation d.”
BackgroundMethodInfluentialVariational Causal Inference Network for Explanatory Visual Question Answering
Xue, D., Qian, S., & Xu, C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00238 · Cited by 27
“We demonstrate the structural causal models (SCMs) [31] of different methods in Figure 3.”
BackgroundMethodNeugebauer, R. S., FIREMAN, B. H., Roy, J. A., Raebel, M. A., Nichols, G. A., & O’Connor, P. J. (2013) · Journal of Clinical Epidemiology · doi:10.1016/j.jclinepi.2013.01.016 · Cited by 28
Causal Asymmetry in a Quantum World
Thompson, J., Garner, A. J. P., Mahoney, J. R., Crutchfield, J. P., Vedral, V., & Gu, M. (2018) · Physical Review X · doi:10.1103/physrevx.8.031013 · Cited by 53
“[56] Judea Pearl, “Causal inference in statistics: An overview,” Statistics surveys 3, 96–146 (2009).”
BackgroundResultBlind Identification of Stochastic Block Models from Dynamical Observations
Schaub, M. T., Segarra, S., & Tsitsiklis, J. N. (2020) · SIAM Journal on Mathematics of Data Science · doi:10.1137/19m1263340 · Cited by 40
“, [5, 37, 53], in which one aims to infer directed graphs that encode direct causal influences between the entities.”
BackgroundData Farming: Methods for the Present, Opportunities for the Future
Sanchez, S. M. (2020) · ACM Transactions on Modeling and Computer Simulation · doi:10.1145/3425398 · Cited by 37
“Causal inference from observational data is a developing area of research (e.g., see [41, 42, 72]), but such analysis often requires strong domain-specific assumptions.”
BackgroundHeckman, J. J., & Pinto, R. R. (2022) · National Bureau of Economic Research · doi:10.3386/w29787 · Cited by 35
“Instead, Pearl (2009b) develops a DAG criteria to check for analogs to matching conditions in the empirical model.”
BackgroundMethodInfluentialThorson, J. T., Anderson, S. C., Goddard, P., & Rooper, C. N. (2025) · Global Ecology and Biogeography · doi:10.1111/geb.70035 · Cited by 34
“By itself, however, the well-known `formula` interface for regression packages is not sufficiently expressive, e.g., to represent dependencies among predictor variables that are relevant during mediation, instrumental variable, or attribution analysis (Pearl 2009).”
Tilden, E. L., & Snowden, J. M. (2018) · Journal of Midwifery & Women s Health · doi:10.1111/jmwh.12710 · Cited by 29
Liu, R., Hao, J., Li, J., Wang, S., Liu, H., Zhou, Z., Delville, M., Cheng, J., Wang, K., & Zhu, X. (2020) · The Journal of Physical Chemistry Letters · doi:10.1021/acs.jpclett.0c02115 · Cited by 28
Adversarial learning for counterfactual fairness
Grari, V., Lamprier, S., & Detyniecki, M. (2022) · Machine Learning · doi:10.1007/s10994-022-06206-8 · Cited by 45
Improving Visualization Interpretation Using Counterfactuals
Kaul, S., Borland, D., Cao, N., & Gotz, D. (2021) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2021.3114779 · Cited by 33
“Many approaches exist to identify and model causal relationships in observational data [44, 53, 59], and several visualization systems support such analyses with suites of data visualization and interaction tools.”
BackgroundImproving Weakly Supervised Object Localization via Causal Intervention
Shao, F., Luo, Y., Zhang, L., Lu, Y., Tang, S., Yang, Y., & Xiao, J. (2021) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3474085.3475485 · Cited by 34
“Causal inference [27, 35, 48] is a critical research topic across many domains, such as statistics [28], politics [14], psychology, and epidemiology [21, 30].”
BackgroundFakhar, K., & Hilgetag, C. C. (2022) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1010250 · Cited by 31
“In contrast to correlational approaches that quantify the statistical association between system elements and functions [5], perturbing and manipulating the system results in changes that can be used to pinpoint actual causes, since the manipulation is assumed to differentiate contributing causes from covarying confou…”
BackgroundFairness without the Sensitive Attribute via Causal Variational Autoencoder
Grari, V., Lamprier, S., & Detyniecki, M. (2022) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2022/98 · Cited by 51
“It aims at building prediction models which ensure fairness at the most individual level, by requiring the notion of the do operator (Pearl et al. 2009) with the intervention of the sensitive in causality ( do ( s = 1) or do ( s = 0) ).”
BackgroundICEEMDAN-Based Transfer Entropy between Global Commodity Classes and African Equities
Bossman, A., & Agyei, S. K. (2022) · Mathematical Problems in Engineering · doi:10.1155/2022/8964989 · Cited by 26
“Schreiber [25] accentuates that TE is a theoretic quantification of information transfer from one market to another.,e philosophical principles of Dretske [26] and the statistics of Pearl [27] facilitate the quantification of the situated information transferred between two markets, as hypothesised by Benthall [28].”
BackgroundCausal deep learning models for studying the Earth system
Tesch, T., Kollet, S. J., & Garcke, J. (2023) · Geoscientific model development · doi:10.5194/gmd-16-2149-2023 · Cited by 32
“suitable additional input variables are included (Pearl, 2009).”
BackgroundMethodInfluentialCarbon dioxide in the soils and adjacent caves of the Moravian Karst
Faimon, J., & Ličbinská, M. (2010) · Acta Carsologica · doi:10.3986/ac.v39i3.76 · Cited by 24
“Spurious relationship problem It is well known t�at statistically related variables (correlated) need not s�ow a causal connection and t�at t�e correlation can be t�e result of a spurious relations�ip (see, e.g., Ben-Zeev & Star 2001; Pearl 2009).”
BackgroundInfluentialSpeicher, P. J., Ganapathi, A. M., Englum, B. R., et al. (2013) · Journal of the American College of Surgeons · doi:10.1016/j.jamcollsurg.2013.12.005 · Cited by 24
“Variables chosen for the matching algorithm were those patient- and disease-related characteristics thought most likely to act as confounders.[18-20] These variables were entered into a logistic regression model to calculate propensity scores, and a nearest-neighbor algorithm was employed to find the most appropriate…”
MethodSeasonal variations of decay rate measurement data and their interpretation
Schrader, H. (2016) · Applied Radiation and Isotopes · doi:10.1016/j.apradiso.2016.05.001 · Cited by 24
Beneficial Action within Altruistic and Prosocial Behavior
Toumbourou, J. W. (2016) · Review of General Psychology · doi:10.1037/gpr0000081 · Cited by 34
“However, evaluation techniques for causal inference have advanced considerably in psychology and other empirical disciplines in recent decades (Cook & Campbell, 1979; Pearl, 2009).”
BackgroundWebb, C. A., Stanton, C. H., Bondy, E., Singleton, P., Pizzagalli, D. A., & Auerbach, R. P. (2019) · Journal of Consulting and Clinical Psychology · doi:10.1037/ccp0000393 · Cited by 25
“Unless relevant confounding between-patient characteristics are modeled, a spurious relationship is possible (but see Pearl (2009) regarding blocking back-door pathways from the mediator to the outcome).”
Integer Programming for Learning Directed Acyclic Graphs from Continuous Data
Manzour, H., Küçükyavuz, S., Wu, H.-H., & Shojaie, A. (2020) · INFORMS Journal on Optimization · doi:10.1287/ijoo.2019.0040 · Cited by 47
Liu, L., Gao, X. A., Zhuang, J., Wen, W., Yang, B., Cheng, W., Xiao, P., Yao, X., & Deng, O. (2020) · Land Use Policy · doi:10.1016/j.landusepol.2020.104623 · Cited by 24
A Survey on Causal InferencePreprint
Yao, L., Chu, Z., Li, S., et al. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.02770 · Cited by 24
“tching based methods [131], survey of tree-based and ensemble-based method [10], and the review of dynamic treatment regimes [25]. For the structural causal model, it is suggested to refer the survey [91] or the book [90]. We will also briefly discuss the relation and difference between the two causality frameworks at…”
BackgroundMethodCausal Navigation by Continuous-time Neural NetworksPreprint
Vorbach, C., Hasani, R., Amini, A., Lechner, M., & Rus, D. L. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.08314 · Cited by 62
“For long, causal modeling frameworks aims at bridging this gap to extract statistical dependencies while constructing causal graphs (Pearl, 2009, Spirtes et al., 2000) or structural causal models(Schölkopf, 2019) to intervene and explain relationships.”
BackgroundMethodCA-SpaceNet: Counterfactual analysis for 6D pose estimation in space
Wang, S., Wang, S., Jiao, B., Yang, D., Su, L., Zhai, P., Chen, C., & Zhang, L. (2022) · 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9981172 · Cited by 24
Das, S. R., Stanton, R., & Wallace, N. E. (2023) · Annual Review of Financial Economics · doi:10.1146/annurev-financial-110921-125930 · Cited by 61
Skowron, A., Jankowski, A. W., & Świniarski, R. W. (2015) · Springer eBooks · doi:10.1007/978-3-662-43505-2_21 · Cited by 23
Yazdani, A., Yazdani, A., Elsea, S. H., Schaid, D. J., Kosorok, M. R., Dangol, G., & Samiei, A. (2019) · BMC Genomics · doi:10.1186/s12864-019-5772-4 · Cited by 26
Analytical Problem Solving Based on Causal, Correlational and Deductive Models
de Mast, J., Steiner, S. H., Nuijten, W., & Kapitan, D. (2022) · The American Statistician · doi:10.1080/00031305.2021.2023633 · Cited by 32
“Observational studies could be used provided that these can be augmented to warrant causal inference, for example, using structural causal modeling (Lederer et al. 2019; Pearl 2009).”
BackgroundMethodInfluentialReisenzein, R., Hildebrandt, A., & Weber, H. (2020) · Cambridge University Press eBooks · doi:10.1017/9781108264822.009 · Cited by 37
Hagedorn, C., Huegle, J., & Schlößer, R. (2022) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-022-01952-x · Cited by 33
“An intervention on V i changes V i to a fixed value v i , which graphically matches the deletion of all incoming edges of V i in G transforming the observational to an experimental setup (Pearl, 2009a).”
BackgroundMethodInfluentialMu, S., Li, Y., Zhao, W. X., Wang, J., Ding, B., & Wen, J.-R. (2022) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3477495.3531934 · Cited by 39
“Following the structural causal model [19, 20], we can formulate the knowledge-aware recommendation in the form of the causal graph [29] in Figure 4.”
MethodFatima, T., & Elbanna, S. (2023) · Tourism Management Perspectives · doi:10.1016/j.tmp.2023.101141 · Cited by 52
Metadata
- Authors
- Pearl, J.
- Affiliations
- University of California, Los Angeles (US)
- Corresponding author
- Pearl, Judea
- Venue
- Statistics Surveys
- Publisher
- Institute of Mathematical Statistics
- Published
- 2009-01-01
- Volume
- 3
- Issue
- none
- Pages
- 96-146
- ISSN
- 1935-7516 (ISSN-L)
- OpenAlex
- W2143117649
- Cited by
- 2,336
- References
- 135
- Citations by year
- 2026: 184 · 2025: 273 · 2024: 271 · 2023: 245 · 2022: 215 · 2021: 208 · 2020: 243 · 2019: 140 · 2018: 96 · 2017: 77 · 2016: 82 · 2015: 82 · 2014: 63 · 2013: 65 · 2012: 46
- FWCI
- 25.47
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Statistics Surveys (publishedVersion, cc-by)
- Project Euclid (Cornell University) (submittedVersion)
- eScholarship (California Digital Library) (submittedVersion)
- Department of Statistics, UCLA
- Funders
- Office of Naval Research
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Modeling and Causal Inference
- Social Sciences › Arts and Humanities › History and Philosophy of Science › Philosophy of Biology and Scientific Explanation
- Keywords
- causal inference, structural causal models, counterfactual reasoning, causal assumptions, causal effects, policy evaluation, intervention effects, regret, causal attribution, direct and indirect effects, mediation analysis
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.