Pith. sign in

REVIEW 2 major objections 4 minor 23 references

Conditional Independence Tests for Constraint-Based Causal Discovery: A Survey

T0 review · 2 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Conditional-independence tests decide every edge in constraint-based causal discovery, and six families of tests come with sharply different failure modes.

desk verdict A genuinely useful CI-test survey with an accurate core synthesis, a real but local technical error about Benjamini-Hochberg, and heuristic guidance that is honestly labeled as unvalidated. read the letter →

arxiv 2608.11156 v1 pith:FJA6NDNS submitted 2026-08-11 stat.ML cs.LG

Pavel Averin , Theodoros Moysiadis , Ioannis Katakis This is my paper · ORCID
classification stat.MLcs.LG MSC 62H2062H1562G10
keywords conditionalindependencetestingconstraint-basedcausaldiscoveryPCalgorithmFCIskeletonrecoveryv-structureorientationmixed-typedatastatisticalpower
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Conditional-independence tests are the statistical engine of constraint-based causal discovery: algorithms such as PC (Peter-Clark), FCI (Fast Causal Inference), and Grow-Shrink remove edges and orient v-structures purely on the outcome of tests of the form $X \perp Y \mid Z$. This survey argues that graph quality is only as good as the validity of those CI decisions, and organizes the field into six test families—partial-correlation, contingency-table, regression, nearest-neighbor, kernel, and machine-learning-based—to make the choice tractable. It connects test-level behavior (power decaying as conditioning sets grow, asymmetric consequences of false positives and false negatives) to graph-level failures (spurious or missing edges, wrong v-structure orientations). It then translates those links into heuristic decision diagrams for continuous, categorical, and mixed-type data, and compares which families are actually available in R and Python software. A reader should care because the same CI test choice that governs every edge also shapes interpretability, scalability, and reliability in high-stakes biomedical applications.

What carries the argument

The load-bearing object is the conditional-independence test itself, treated as the engine of skeleton pruning and orientation. The paper's organizing machinery is a six-family taxonomy—partial-correlation, contingency-table, regression, nearest-neighbor, kernel, and machine-learning-based—augmented by robustness layers (shrinkage, permutation, test aggregation) that wrap base tests without becoming separate families. The mechanism that carries the argument from tests to graphs is the decision chain: a test of $X \perp Y \mid Z$ removes or keeps an edge, the conditioning set that yields independence becomes the separating set for that edge, separating sets determine whether an unshielded triple is oriented as a v-structure, and orientations propagate through Meek's rules into a completed partially directed acyclic graph or partial ancestral graph. Within that chain, power decay of every family as $|Z|$ grows, and the asymmetric impact of type I versus type II errors, are the quantities that explain graph-level failure modes.

What would settle it

A controlled benchmark holding the discovery algorithm fixed and varying only the CI-test family, sample size, conditioning-set depth, and graph density, on standard benchmark networks and semi-synthetic mixed-type data, would settle the central link: if per-test calibration and power rankings stop predicting graph-level skeleton and orientation accuracy—for example, a better-calibrated test producing worse graphs—the survey's core claim about CI tests as the decisive engine would fail. Section 10.5 notes that exactly this experiment is still missing.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that conditional-independence testing, not scoring or global optimization, is the decisive statistical component of constraint-based causal discovery. In PC, FCI, and related algorithms, a CI test outcome either keeps or deletes an edge, stores the separating set, and later sets the orientation of unshielded triples; those orientations then propagate through orientation rules into the final completed partially directed acyclic graph or partial ancestral graph. The survey's central assertion is that the validity of the learned graph is only as good as the validity of these CI decisions, and that test selection is the main lever a practitioner controls. It reviews six families of tests, states for each when its decisions reflect the data-generating distribution and when they fail, and shows that errors propagate asymmetrically: false positives add spurious edges, while false negatives delete true edges and can destroy separating sets needed for correct v-structure orientation, with cascading effects through subsequent orientation rules. It closes with heuristic selection frameworks and an explicit list of open problems, mixed-type testing without discretization and small-sample error control among them.

Load-bearing premise

The guidance assumes that the strengths and weaknesses of each test family, measured in isolation, still hold when the same test is embedded in an iterative PC- or FCI-style search across real mixed-type datasets—an assumption the paper itself states has not been systematically validated.

Editorial extensions

If this is right

  • Any change in CI-test calibration propagates directly into the graph: inflated false positives add spurious edges, and inflated false negatives remove true edges and can erase the separating sets that later v-structure orientation depends on.
  • Because power decays with conditioning-set size for every family, capping $|Z|$ or switching to Markov-blanket and local discovery methods is the practical safeguard when the sample size is small relative to graph degree.
  • For continuous Gaussian-like data Fisher's $Z$ is the fastest and best-powered default; for categorical data with small-to-moderate conditioning sets $G^2/\chi^2$ tests are effective; for nonlinear continuous data the choice hinges on budget and conditioning depth (KNN-CMI for small $|Z|$, RCIT/RCoT for large $|Z|$, full KCI only with ample compute).
  • No major library surveyed ships a native machine-learning-based CI test, so DML-style and generative tests are currently research tools rather than drop-in options for practitioners.
  • Reporting the CI test, its significance level, the maximum conditioning-set size, and key separating sets should become standard practice, because the audit trail of a constraint-based graph is exactly the sequence of CI decisions.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's Section 10.5 concedes that the decision diagrams in Figures 6–8 have not been validated by a systematic empirical comparison; the natural next step is a controlled benchmark that plugs each family into the same PC/FCI implementation while varying sample size, conditioning depth, and graph density.
  • The asymmetric-error argument suggests that safety-critical applications should tune $\alpha$ toward avoiding false negatives rather than toward global false-positive control, because type II errors cascade through orientation—a consequence the survey notes but does not quantify.
  • If power decay with $|Z|$ is as steep as the survey describes, algorithmic restructuring that reduces conditioning depth (local Markov-blanket search, divide-and-conquer skeleton learning) could improve graph accuracy even without improving per-test power.
  • The gap between methodological development and tooling implies that near-term practical gains are more likely to come from robust wrappers around existing tests (permutation, wild bootstrap, aggregation) than from entirely new CI tests.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. This survey treats conditional independence testing as the core statistical engine of constraint-based causal discovery, arguing that the validity of CI decisions controls skeleton recovery and edge orientation in algorithms such as PC and FCI. It organizes the literature into six families—partial-correlation, contingency-table, regression, nearest-neighbor, kernel, and machine-learning-based—and discusses their assumptions, calibration, computational cost, and robustness layers. It also links test-level properties such as power decay with conditioning-set size and asymmetric type I/II error costs to graph-level failure modes, compares software support across R and Python libraries, and closes with practitioner-oriented decision diagrams and open problems. The paper is a synthesis rather than a new methodological contribution.

Significance. If the issues below are addressed, this survey would be a genuinely useful reference for entering researchers and practitioners. Its strengths are a clear family taxonomy, compact and mostly accurate descriptions of each CI test with appropriate citations, detailed software comparison tables, and a candid treatment of open problems—including, notably, the admission in Section 10.5 that the survey's own decision heuristics are not yet validated. The organizing theme that per-test CI behavior propagates to graph-level errors is valuable and gives the survey a coherent message. The paper does not introduce fitted parameters or new entities, and its claims are summaries of published results, so circularity is not a concern.

major comments (2)
  1. [Section 6.2.5] The statement that 'Benjamini–Hochberg, meanwhile, controls the FDR only for independent test statistics' is incorrect. The Benjamini–Hochberg procedure controls the FDR under positive regression dependence on the subset of true null hypotheses, and under arbitrary dependence it still controls the FDR at level (m0/m) times the nominal level. The true issue for PC is whether the collection of CI test statistics satisfies the required dependence conditions, not that BH requires independence. This paragraph should be corrected and the motivation for recommending Benjamini–Yekutieli rephrased accordingly.
  2. [Section 2 / Figures 6–8 / Section 10.5] The practitioner decision frameworks in Figures 6–8 are presented as contribution 4, but the manuscript itself concedes in Section 10.5 that a systematic empirical comparison across data types, sample sizes, conditioning-set depths, and graph densities is still lacking, and that the proposed heuristics 'need confirmation or refinement.' This matters because Sections 6.2.2 and 6.2.5 explain that type I/II errors propagate asymmetrically through iterative discovery and that repeated adaptive testing changes the operating characteristics of a test; therefore, per-test calibration and power do not automatically transfer to graph-level performance in PC or FCI. I recommend either clearly reframing Figures 6–8 as provisional hypotheses with the Section 10.5 caveat repeated in Section 6.3 and the figure captions, or adding a focused validation study. As written, the practical guidance is stronger than the evidence presented.
minor comments (4)
  1. [Section 5.2] The degrees-of-freedom expression for G2 and chi-square is ambiguous as typeset: it should explicitly sum over the strata of the conditioning variable Z and clarify whether the category counts r_x and r_y are stratum-dependent.
  2. [Figure 6 caption] The claim that 'empirical evidence suggests that KNN-CMI is better calibrated at small |Z|, while RCoT/RCIT achieve higher power at large |Z|' needs a precise pointer to the specific experiments in Runge (2018); as written, the claim is not checkable.
  3. [Section 10.2] Saying that small-sample CI tests can exhibit 'inflated false discovery rates' is imprecise; the per-test quantity that inflates is the false positive rate, whereas FDR is a property of a multiple-testing procedure.
  4. [Section 6.3] The decision diagrams use qualitative labels such as 'Small', 'Med.', and 'Large' without any quantitative anchors; adding rough sample-size or conditioning-depth thresholds would make the guidance more actionable.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the survey summarizes an independent literature and explicitly flags its own heuristics as unvalidated rather than presenting them as derived predictions.

full rationale

This is a survey paper with no derivation chain, no fitted parameters, and no new entities whose properties are defined in terms of the claims. The central assertion that CI tests are the statistical engine of constraint-based discovery and that graph quality depends on CI validity is a synthesis of the published PC/FCI literature (Spirtes et al., 2000; Kalisch & Bühlmann, 2007; Colombo & Maathuis, 2014), which is external and independent of the present authors. The practitioner decision frameworks in Figures 6–8 are explicitly presented as heuristics grounded in theoretical properties and individual empirical findings; Section 10.5 concedes that “a systematic empirical comparison of CI test families across data types, sample sizes, conditioning set depths, and graph densities is still lacking” and that the proposed heuristics “need confirmation or refinement.” That is an honest statement of an open empirical gap, not a circular reduction of a prediction to its inputs. The authors’ own prior papers (Ganopoulou et al., 2024; 2025; Boutsika et al., 2023; Skodra et al., 2023) appear only as biomedical application examples and do not carry the methodological argument. No step in the paper reduces by construction to its own inputs, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 0 assumptions · 0 invented entities

This is a survey paper. It introduces no free parameters, postulates no new entities, and relies on the standard assumptions of the cited literature (Markov condition, faithfulness, causal sufficiency) only as subjects of review, not as premises of a new derivation. The self-cited biomedical papers are used as illustrative applications and do not ground the survey's methodological claims.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Conditional Independence Tests for Constraint-Based Causal Discovery: A Survey." pith.science (2026). https://pith.science/paper/FJA6NDNS

@misc{pith2026260811156,
  author       = {Pith},
  title        = {Pith review of: Conditional Independence Tests for Constraint-Based Causal Discovery: A Survey},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FJA6NDNS}},
  note         = {Machine review of arXiv:2608.11156}
}
read the original abstract

Conditional Independence (CI) tests are the statistical engine of constraint-based causal discovery: in algorithms such as PC (Peter-Clark) and FCI (Fast Causal Inference), skeleton pruning and key orientations follow directly from CI decisions. This survey reviews CI testing with emphasis on assumptions, robustness, and scalability in high-dimensional and mixed-type settings common in biomedical domains. The survey organizes widely used CI methods into six families: partial-correlation, contingency-table, regression, nearest-neighbor, kernel, and machine-learning-based. Special emphasis is provided on the robustness layers that address the limitations of these families. For each family, the survey examines when CI decisions reflect the data-generating distribution and when they fail. By this, we link test-level properties, including power decay with conditioning set size and asymmetric type I/II error consequences, to graph-level errors in skeleton recovery and v-structure orientation. The survey also compares adoption across major R and Python libraries and summarizes open challenges, including mixed-type CI testing without discretization, small-sample error control, and strategies for improving scalability of CI-testing.

Figures

Figures reproduced from arXiv: 2608.11156 by the authors.

Figure 1
Figure 1. illustrates the Markov blanket of a node X: once the values of its parents, children, and spouses are known, X is conditionally independent of all remaining variables in the graph. X P1 P2 C1 C2 S O1 O2 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. DAG illustrating the Causal Markov Condition. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. DAG illustrating a Faithfulness violation via path cancellation. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: CPDAG inferred when L is unmea￾sured: under (incorrectly assumed) causal suffi￾ciency the dependence between Y and Z yields an unoriented edge Y − Z. x 6 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Heuristic decision diagram for selecting CI tests for continuous variables. The conditioning set [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Heuristic CI test selection for categorical variables. For small-to-moderate conditioning sets, [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: Heuristic CI test selection for mixed-type data. The small-budget discretization path is viable only [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 16 canonical work pages

  1. [14]

    URLhttps://ojs

    doi: 10.1609/aaai.v39i18.34062. URLhttps://ojs. aaai.org/index.php/AAAI/article/view/34062. Michael I. Love, Wolfgang Huber, and Simon Anders. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.Genome Biology, 15(12):550,

    work page doi:10.1609/aaai.v39i18.34062

  2. [15]

    Daniel Marbach, Robert J Prill, Thomas Schaffter, Claudio Mattiussi, Dario Floreano, and Gustavo Stolovitzky

    doi: 10.1186/s13059-014-0550-8. Daniel Marbach, Robert J Prill, Thomas Schaffter, Claudio Mattiussi, Dario Floreano, and Gustavo Stolovitzky. Revealing strengths and weaknesses of methods for gene network inference.Proceedings of the national academy of sciences, 107(14):6286–6291,

    doi:10.1186/s13059-014-0550-8

  3. [22]

    Strobl, Peter L

    Eric V. Strobl, Peter L. Spirtes, and Shyam Visweswaran. Estimating and controlling the false discovery rate of the PC algorithm using edge-specific P-values.ACM Transactions on Intelligent Systems and Technology, 10(5):1–37, 2019a. doi: 10.1145/3351342. Eric V Strobl, Kun Zhang, and Shyam Visweswaran. Approximate kernel-based conditional independence tes...

    work page doi:10.1145/3351342

  4. [23]

    Ioannis Tsamardinos and Laura E Brown

    doi: 10.1007/s41060-018-0097-y. Ioannis Tsamardinos and Laura E Brown. Bounding the false discovery rate in local Bayesian network learning. InAAAI, pp. 1100–1105,

    work page doi:10.1007/s41060-018-0097-y

  5. [24]

    gcastle: A Python toolbox for causal discovery.arXiv preprint arXiv:2111.15155,

    32 Published in Transactions on Machine Learning Research (07/2026) Keli Zhang, Shengyu Zhu, Marcus Kalander, Ignavier Ng, Junjian Ye, Zhitang Chen, and Lujia Pan. gcastle: A Python toolbox for causal discovery.arXiv preprint arXiv:2111.15155,

    arXiv 2026

  6. [1953]

    Causal discovery from heterogeneous/nonstationary data.Journal of Machine Learning Research, 21(89):1–53,

    28 Published in Transactions on Machine Learning Research (07/2026) Biwei Huang, Kun Zhang, Jiji Zhang, Joseph Ramsey, Ruben Sanchez-Romero, Clark Glymour, and Bern- hard Schölkopf. Causal discovery from heterogeneous/nonstationary data.Journal of Machine Learning Research, 21(89):1–53,

    work page 2026

  7. [1961]

    27 Published in Transactions on Machine Learning Research (07/2026) Amir Foroushani, Rupesh Agrahari, Roderick Docking, Linda Chang, Gerben Duns, Monika Hudoba, Aly Karsan, and Habil Zare. Large-scale gene network analysis reveals the significance of extracellular matrix pathway and homeobox genes in acute myeloid leukemia: an introduction to the Pigengen...

    work page 2026

  8. [1987]

    Peter Spirtes and Clark Glymour

    Reprint of the 1904 article. Peter Spirtes and Clark Glymour. An algorithm for fast recovery of sparse causal graphs.Social science computer review, 9(1):62–72,

    work page 1904

Show all 23 references
  1. [2000]

    Disclosing the molecular basis of salinity priming in olive trees using proteogenomic model discovery.Plant Physiology, 191(3):1913–1933,

    Christina Skodra, Michail Michailidis, Theodoros Moysiadis, George Stamatakis, Maria Ganopoulou, Ioannis-Dimosthenis S Adamakis, Lefteris Angelis, Ioannis Ganopoulos, Georgia Tanou, and Martina Samiotaki. Disclosing the molecular basis of salinity priming in olive trees using ...

    1913

  2. [2002]

    doi: 10.1002/0471249688

    ISBN 9780471360933. doi: 10.1002/0471249688. Constantin F Aliferis, Ioannis Tsamardinos, and Alexander Statnikov. HITON: a novel Markov blanket algorithm for optimal variable selection. InAMIA annual symposium proceedings, volume 2003, pp. 21,

    doi

  3. [2004]

    LLM-driven causal discovery via harmonized prior.IEEE Transactions on Knowledge and Data Engineering, 37(4):1943–1960,

    Taiyu Ban, Lyuzhou Chen, Derui Lyu, Xiangyu Wang, Qinrui Zhu, and Huanhuan Chen. LLM-driven causal discovery via harmonized prior.IEEE Transactions on Knowledge and Data Engineering, 37(4):1943–1960,

    1943

  4. [2008]

    Causal discovery with continuous additive noise models.Journal of Machine Learning Research, 15:2009–2053,

    Jonas Peters, Joris M Mooij, Dominik Janzing, and Bernhard Schölkopf. Causal discovery with continuous additive noise models.Journal of Machine Learning Research, 15:2009–2053,

    2009

  5. [2010]

    Learning Bayesian network model structure from data

    29 Published in Transactions on Machine Learning Research (07/2026) Dimitris Margaritis. Learning Bayesian network model structure from data. Technical report,

    2026

  6. [2013]

    doi: 10.1002/9781118548387

    ISBN 9780470582473. doi: 10.1002/9781118548387. Harold Hotelling. New light on the correlation coefficient and its transforms.Journal of the Royal Statistical Society. Series B (Methodological), 15(2):193–232,

    doi

  7. [2014]

    Efficient ensemble conditional independence test framework for causal discovery.arXiv preprint arXiv:2509.21021,

    Zhengkang Guan and Kun Kuang. Efficient ensemble conditional independence test framework for causal discovery.arXiv preprint arXiv:2509.21021,

  8. [2016]

    Permutation-based rank test in the presence of discretization and application in causal discovery with mixed data.arXiv preprint arXiv:2501.18990,

    Xinshuai Dong, Ignavier Ng, Boyang Sun, Haoyue Dai, Guang-Yuan Hao, Shunxing Fan, Peter Spirtes, Yumou Qiu, and Kun Zhang. Permutation-based rank test in the presence of discretization and application in causal discovery with mixed data.arXiv preprint arXiv:2501.18990,

    arXiv

  9. [2017]

    Mixed graphical models for causal analysis of multi-modal variables.arXiv preprint arXiv:1704.02621,

    Andrew J Sedgewick, Joseph D Ramsey, Peter Spirtes, Clark Glymour, and Panayiotis V Benos. Mixed graphical models for causal analysis of multi-modal variables.arXiv preprint arXiv:1704.02621,

    arXiv

  10. [2018]

    William G Cochran

    doi: 10.1111/ectj.12097. William G Cochran. Some methods for strengthening the commonχ2 tests.Biometrics, 10(4):417–451,

    doi

  11. [2019]

    Adjacency-faithfulness and conservative causal inference

    30 Published in Transactions on Machine Learning Research (07/2026) Joseph Ramsey, Jiji Zhang, and Peter Spirtes. Adjacency-faithfulness and conservative causal inference. In Proceedings of the Twenty-Second Conference on Uncertainty in Artificial Intelligence (UAI), pp. 401–408,

    2026

  12. [2022]

    Partial correlation and conditional correlation as mea- sures of conditional independence.Australian & New Zealand Journal of Statistics, 46(4):657–664,

    26 Published in Transactions on Machine Learning Research (07/2026) Kunihiro Baba, Ritei Shibata, and Masaaki Sibuya. Partial correlation and conditional correlation as mea- sures of conditional independence.Australian & New Zealand Journal of Statistics, 46(4):657–664,

    2026

  13. [2023]

    A new coefficient of correlation.Journal of the American Statistical Association, 116 (536):2009–2022,

    Sourav Chatterjee. A new coefficient of correlation.Journal of the American Statistical Association, 116 (536):2009–2022,

    2009

  14. [2024]

    Thuc Duy Le, Tao Hoang, Jiuyong Li, Lin Liu, Huawen Liu, and Shu Hu

    URLhttps://arxiv.org/abs/2407.15073. Thuc Duy Le, Tao Hoang, Jiuyong Li, Lin Liu, Huawen Liu, and Shu Hu. A fast PC algorithm for high dimensional causal discovery with multi-core PCs.IEEE/ACM transactions on computational biology and bioinformatics, 16(5):1483–1495,

    arXiv

  15. [2025]

    URLhttps://doi.org/10.1109/TKDE.2025.3528461

    doi: 10.1109/TKDE.2025.3528461. URLhttps://doi.org/10.1109/TKDE.2025.3528461. Ingo A Beinlich, Henri Jacques Suermondt, R Martin Chavez, and Gregory F Cooper. The ALARM mon- itoring system: A case study with two probabilistic inference techniques for belief networks. InAIME 89...

    2025

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.