The DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index; 3 of the 7 sources its claims cite were checked for support.
Confounding Equivalence in Causal Inference
Open accessAbstract
The paper provides a simple test for deciding, from a given causal diagram, whether two sets of variables have the same bias-reducing potential under adjustment. The test requires that one of the following two conditions holds: either (1) both sets are admissible (i.e. satisfy the back-door criterion) or (2) the Markov boundaries surrounding the treatment variable are identical in both sets. We further extend the test to include treatment-dependent covariates by broadening the back-door criterion and establishing equivalence of adjustment under selection bias conditions. Applications to covariate selection and model testing are discussed.
Overview
CoChat Summary
TL;DR
The paper gives a graphical test for deciding whether two covariate sets have the same bias-reducing potential under adjustment, called c-equivalence. It shows that, for sets containing no descendants of the treatment, c-equivalence holds iff the sets are both back-door admissible or have identical Markov boundaries.
Key findings
- Two sets \(T\) and \(Z\) are c-equivalent if their adjustment estimands are equal for every \(x,y\), meaning they would produce the same asymptotic bias relative to the target causal effect.
- Theorem 2 states a necessary and sufficient condition: for sets containing no descendant of \(X\), \(T\) and \(Z\) are c-equivalent iff either \(Z_m = T_m\) (identical Markov boundaries) or both \(Z\) and \(T\) are \(G\)-admissible by the back-door criterion.
- Theorem 1 and Corollary 1 give sufficient statistical conditions, \(C^*\): \(X \perp Z \mid T\) and \(Y \perp T \mid Z,X\); \(C^{}\): \(X \perp T \mid Z\) and \(Y \perp Z \mid T,X\), but these do not characterize all c-equivalent pairs.
- The paper reports that c-equivalence and \(G\)-admissibility can be tested in polynomial time, and that failing a c-equivalence test can rule out models in which two sets are both admissible.
Methods
- Theoretical causal-inference and graphical-methods paper using directed acyclic graphs, d-separation, the back-door criterion, admissibility, and Markov boundaries.
- The analysis consists of definitions, theorems, corollaries, and mathematical proofs, with illustrative DAG examples; no empirical data, sample, simulation, or numerical estimates are reported.
- Appendices prove uniqueness of the Markov boundary, prove monotonicity of d-separation over minimal subsets, and discuss propensity-score dimensionality reduction for c-equivalence testing.
Limitations
- Stated: Theorem 1 and Corollary 1 do not characterize all c-equivalent pairs; the paper gives Figure 1 as an example where both sets are \(G\)-admissible but neither \(C^*\) nor \(C^{}\) holds.
- Stated: Theorem 2’s necessary part excludes descendants of \(X\); a footnote says sufficiency holds without excluding descendants, and necessity can allow descendants if \(Z_m\) and \(T_m\) are interpreted as reduced Markov boundaries.
- Stated: no statistical test can confirm or refute the admissibility of a given set \(Z\); c-equivalence tests can only show that two sets cannot both be admissible. The paper also notes it is an open question whether c-equivalence has the same discrimination power as conditional-independence tests. Evident: the work is purely theoretical and offers no empirical validation.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
3 Valid · 0 Not valid · 3 Not checked · 15 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“It is well known (Rosenbaum and Rubin, 1983) that, viewed as a random variable, L ( z ) satisfies X ⊥ ⊥ L ( z ) | Z . This implies that A ( x, y, L ( z )) = A ( x, y, Z ) and, therefore, testing for the c -equivalence of Z and T can be reduced to testing the c -equivalence of L ( z ) and L ( t ).”
What the sources say
- Supportedcites Rosenbaum and Rubin, 1983 source
Theorem 1 in §2-2 (page not provided) states that treatment assignment and observed covariates are conditionally independent given the propensity score. That supports the result explicitly attributed to Rosenbaum and Rubin in the citing sentence; the subsequent reduction to testing c-equivalence is presented as the citing authors’ own inference, not as a result stated in the source. (LLM, conf=0.90)
boosted- 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
“To test whether ( X ⊥ ⊥ Y | Z ) holds in a distribution P , it is necessary that ( X ⊥ ⊥ Y | L ( z )) holds in P .”
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.
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
“It follows from the proof above that the Markov boundary S m is the intersection of all subsets of S that d -separate S from X .”
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.
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
“While perfectly valid, this condition is still not complete; there are cases where T and Z are c -equivalent, yet none is c -equivalent to their union.”
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.
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
“Lemma 2. Every set of variables, S , is c -equivalent to its Markov boundary S m .”
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.
11 more claims
Create a free account to see all 21 claims and every source verdict.
- Probably not a substantial claim. 1
- It may state two separate claims as one. 4
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?
In the citing paper
“The use of a DAG to specify causal assumptions can aid the identification of potential confounders, as variables blocking back-door paths from exposure to outcome.20”
From A structured approach to hypotheses involving continuous exposures over the life course (2016) · cites this paper as [20]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 4 STEP-WISE TEST FOR c -EQUIVALENCE
“Figure 2: W 3 and W 4 are non-admissible yet c -equivalent; both having ∅ as a Markov boundary.”
§ 4 STEP-WISE TEST FOR c -EQUIVALENCE
“and would be unable to proceed toward Z m = { W 3 } .”
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
Causality: Models, Reasoning, and Inference
Pearl, J. (2009) · doi:10.1017/cbo9780511803161 · Cited by 11,653
Self-citationEstimating 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
Discrete 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
Causation, Prediction, and Search
Spirtes, P., Glymour, C., & Scheines, R. (2001) · doi:10.7551/mitpress/1754.001.0001 · Cited by 4,521
Causal Diagrams for Epidemiologic Research
Greenland, S., Pearl, J., & Robins, J. M. (1999) · Epidemiology · doi:10.1097/00001648-199901000-00008 · Cited by 3,728
Self-citationProbabilistic Reasoning in Intelligent Systems
(1988) · Elsevier eBooks · doi:10.1016/c2009-0-27609-4 · Cited by 3,321
Air Pollution and Hospital Admissions for Heart Disease in Eight U.S. Counties
Schwartz, J. (1999) · Epidemiology · doi:10.1097/00001648-199901000-00005 · Cited by 3,038
Causal diagrams for empirical research
PEARL, J. (1995) · Biometrika · doi:10.1093/biomet/82.4.669 · Cited by 2,355 · cited 1×
“Such relationships can effectively be represented in the form of directed acyclic graphs (DAG’s) (Pearl, 1995; Lauritzen, 2001; Spirtes et al., 2000; Glymour and Greenland, 2008; Dawid, 2002).”
Self-citationMaternal Pesticide Exposure from Multiple Sources and Selected Congenital Anomalies
Shaw, G. M., Wasserman, C. R., OʼMalley, C. D., Nelson, V., & Jackson, R. J. (1999) · Epidemiology · doi:10.1097/00001648-199901000-00011 · Cited by 1,784
Research works printed but not linked: 6
- 2008, Modern epidemiology, 183
- 1993, Comment: graphical models, causality, and intervention, Stat Sci, 8, 266
- 2004, Uncertainty in artificial intelligence, proceedings of the twentieth conference, 333
- 2012, Proceedings of the twenty-sixth AAAI conference on artificial intelligence, 698
- 2010, Proceedings of the twenty-sixth conference on uncertainty in artificial intelligence, 527
- 1997, Latent variable modeling and applications to causality, 69
Cited by
Causal inference and the data-fusion problem
Bareinboim, E., & Pearl, J. (2016) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1510507113 · Cited by 621
Self-citationComment: understanding Simpson’s paradox
Pearl, J. (2014) · The American Statistician · doi:10.1080/00031305.2014.876829 · Cited by 206
Self-citationA structured approach to hypotheses involving continuous exposures over the life course
Smith, A. D. A. C., Hardy, R., Heron, J., Joinson, C. J., Lawlor, D. A., Macdonald-Wallis, C., & Tilling, K. (2016) · International Journal of Epidemiology · doi:10.1093/ije/dyw164 · Cited by 59
The Deductive Approach to Causal Inference
Pearl, J. (2014) · Journal of Causal Inference · doi:10.1515/jci-2014-0016 · Cited by 28
Self-citationKuroki, M. (2015) · Bernoulli · doi:10.3150/14-bej664 · Cited by 2
Metadata
- Authors
- Pearl, J., & Paz, A.
- Affiliations
- University of California, Los Angeles (US)
- Technion – Israel Institute of Technology (IL)
- ORCID
- Paz, Azaria
- Venue
- Journal of Causal Inference
- Publisher
- Walter de Gruyter GmbH
- Published
- 2014-03-01
- Volume
- 2
- Issue
- 1
- Pages
- 75-93
- ISSN
- 2193-3677 (ISSN-L)
- 2193-3685
- OpenAlex
- W2096880229
- Cited by
- 35
- References
- 23
- Citations by year
- 2026: 4 · 2025: 4 · 2024: 5 · 2023: 2 · 2022: 2 · 2021: 1 · 2020: 3 · 2017: 1 · 2016: 3 · 2015: 5 · 2014: 4
- FWCI
- 4.30
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Journal of Causal Inference (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- eScholarship (California Digital Library) (submittedVersion)
- arXiv (Cornell University)
- Uncertainty in Artificial Intelligence
- Department of Statistics, UCLA
- Funders
- National Science Foundation (1302448, 0914211, 1249822)
- National Institutes of Health (1r01lm009961-01)
- Office of Naval Research (N00014-10-1-0933, N00014)
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Modeling and Causal Inference
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Inference
- Keywords
- adjustment sets, admissibility, selection bias, covariate selection, model testing, causal graphs, bias mitigation
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.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
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
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Confounding Equivalence in Causal Inference. https://app.cochat.ai/papers/10.1515/jci-2013-0020
- The paper
- Pearl, J., & Paz, A. (2014). Confounding Equivalence in Causal Inference. Journal of Causal Inference. https://doi.org/10.1515/jci-2013-0020
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.