Pith. sign in

REVIEW 3 major objections 6 minor 75 references

Overview and practical recommendations on using Shapley Values for identifying predictive biomarkers via CATE modeling

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper proposes and benchmarks a surrogate-model strategy: fit an XGBoost model to the estimated conditional average treatment effect and read SHAP values off that model, giving one protocol for ranking predictive biomarkers that works…

desk verdict A practical, well-benchmarked tutorial on surrogate SHAP for predictive biomarker discovery, with a real but manageable gap: the surrogate's fidelity is asserted for irreducible learners, not directly tested. read the letter →

arxiv 2505.01145 v1 pith:V7RUEHOQ submitted 2025-05-02 stat.ME stat.ML

David Svensson , Erik Hermansson , Nikolaos Nikolaou , Konstantinos Sechidis , Ilya Lipkovich This is my paper · ORCID
classification stat.MEstat.ML
keywords treatmenteffectheterogeneityconditionalaveragepredictivebiomarkersprognosticSHAPvaluesmeta-learnerscausalforestsurrogatemodel
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

This paper tries to establish a single, practical protocol for turning any CATE estimator into a SHAP-based ranking of predictive biomarkers. The proposed protocol fits an XGBoost surrogate model to the estimated CATE and derives TreeSHAP values from that surrogate, sidestepping the computational blow-up of model-agnostic SHAP and the ambiguity of where to apply SHAP inside multi-stage meta-learners. Through simulations the paper shows the protocol works across T-, S-, X-, R-, DR-learners and Causal Forest, and that the choice of learner materially changes discovery accuracy. The authors also show that global rankings and local SHAP profiles can be used, with caveats, to recover the shape and direction of true treatment-effect modifiers.

What carries the argument

The central object is the surrogate model $M^{(2)}$: a second-stage XGBoost model regressing the estimated CATE, $\hat{\tau}(x)$, on all baseline covariates $x_1,\ldots,x_p$, from which exact TreeSHAP values are computed. It carries the argument by converting any multi-stage CATE estimator—reducible or irreducible—into a single supervised model whose SHAP values are well-defined and cheap to compute, bypassing the exponential coalition enumeration of model-agnostic KernelSHAP.

What would settle it

A concrete check would be a simulation with a known CATE function and a deliberately misspecified or underpowered CATE estimator, comparing surrogate SHAP rankings with oracle rankings from the true $\tau(x)$: if the surrogate ranks a known null covariate above a true modifier in a large-n setting more often than chance, the preservation premise fails.

Watch

Extended reading notes

Core claim

The central claim is that the surrogate approach—fitting an XGBoost model to CATE estimates against all baseline covariates and deriving SHAP values from that model—provides a unified, computationally feasible route to SHAP-based biomarker discovery for any CATE estimator. The paper argues this is valid because the surrogate explains a model that represents estimated CATE as a function of covariates, and it verifies empirically that for reducible learners like R- and DR-learning the surrogate performs no worse than deriving SHAP directly from the final pseudo-outcome regression. Simulation benchmarks then show that the choice of CATE learner strongly affects biomarker discovery: in randomized trials the S-learner performs best, while in observational settings the R-learner dominates; T-learning and Causal Forest perform poorly and can fall below random guessing when prognostic strength is high.

Load-bearing premise

The load-bearing premise is that the surrogate XGBoost model fitted to the estimated CATE preserves the covariate-importance structure of the underlying CATE estimator, so that TreeSHAP values from the surrogate are valid proxies for true predictive biomarker effects; the paper states SHAP explains the model, not the truth, and does not directly verify the premise for irreducible learners.

Editorial extensions

If this is right

  • A single SHAP protocol can replace method-specific importance measures across T-, S-, X-, R-, DR-learners and Causal Forest.
  • Practitioners selecting a CATE learner for biomarker discovery should expect different rankings: S-learner in RCTs, R-learner in observational data.
  • Model-agnostic KernelSHAP becomes impractical beyond about 8–15 covariates, while TreeSHAP on the surrogate remains feasible; the surrogate avoids this bottleneck.
  • For R- and DR-learners, deriving SHAP from the surrogate matches the direct pseudo-outcome approach, so the extra modeling step costs no discovery accuracy.
  • Global SHAP rankings and local SHAP profiles can recover qualitative features of true effect modifiers, but inspecting individual profiles risks false discoveries unless a well-performing learner is used.

Reading between the lines

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

  • The surrogate's validity rests on the surrogate faithfully preserving the CATE estimator's covariate-importance structure; a direct check would be to compare surrogate SHAP rankings against oracle rankings for irreducible learners across varied data-generating processes, which the paper only partially covers.
  • The protocol could be extended to loss-based or interaction-aware SHAP definitions, which would address the paper's own caveat that SHAP explains the model, not the truth.
  • Because the paper fixes the surrogate to XGBoost, the protocol's performance may depend on that choice; testing alternative tree ensembles or other surrogates would clarify how much of the result is the surrogate versus the underlying CATE estimator.
  • The observed failure of T-learning and Causal Forest suggests a diagnostic opportunity: a large gap between surrogate SHAP margins and CATE estimation accuracy could flag learners that confound prognostic with predictive signal.
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

3 major / 6 minor

Summary. The paper develops and benchmarks a 'surrogate SHAP' strategy for identifying predictive biomarkers from CATE models. After reviewing Shapley values, SHAP, and meta-learners, it proposes fitting an XGBoost model to estimated CATEs and deriving TreeSHAP values from that surrogate (Strategy 3), claiming this is a unified, computationally feasible route for any CATE estimator. Simulations under an RCT model (S2) and an observational model (S3) compare T-, S-, X-, R-, DR-learners and Causal Forest using TOP1, NET3, and MARGIN metrics; findings include S-learner best in RCTs, R/DR best in observational data, and poor performance of T-learner and Causal Forest. Additional analyses examine prognostic strength, Strategy 2 versus Strategy 3 for R/DR learners, local SHAP recovery of true effect-modifier functions, and two case studies.

Significance. If the surrogate strategy is valid, it gives practitioners a single protocol for biomarker ranking across heterogeneous CATE learners and avoids the computational burden of KernelSHAP. The paper's strengths include carefully described simulation setups with 100 to 500 iterations and standard errors, reproducible R code, a head-to-head comparison of SHAP against Causal Forest VIP, and falsifiable empirical claims about which learners rank predictive biomarkers well. The central caveat is that the surrogate's fidelity to the underlying CATE learner is asserted rather than directly demonstrated for irreducible learners; the cross-learner conclusions rest on that unverified link.

major comments (3)
  1. [Section 2.4 (Strategy 3); Section 4.3, Table 3] The central claim of the paper, that Strategy 3 provides a valid unified SHAP-based biomarker ranking for any CATE learner, requires that the XGBoost surrogate M(2) preserve the covariate-importance structure of tau_hat. The manuscript verifies this only for the reducible R- and DR-learners (Section 4.6) and in a single T-learner toy with p=5 (Section 3.2); no direct fidelity check is reported for S-, T-, X-, or Causal Forest learners in the p=19-20 benchmark of Section 4.3. Since M(2) is fitted in-sample and TreeSHAP is computed on the same training data, overfitting of noise in tau_hat could make the resulting rankings surrogate artifacts rather than properties of the underlying learner. I therefore recommend adding a validation experiment that, for each learner, reports the out-of-sample fidelity of M(2) to tau_hat (e.g., held-out R^2 or rank correlation between M(2) and tau_hat), a small-p (p <= 8) comparison of Strategy 3 SHAP rankings with direct strategy-appropriate SHAP rankings from the learner itself, and sensitivity of TOP1, NET3, and MARGIN to XGBoost hyperparameter choices. Without such a check, the headline conclusions about learner choice remain statements about the surrogate pipeline.
  2. [Section 4.6, Figure 8] The claim that 'no empirical evidence was found to indicate a statistical difference in performance between the two types of SHAPs' is currently supported only by visual inspection of boxplots; no formal test or equivalence margin is reported. Because this Strategy 2 versus Strategy 3 comparison is the main direct evidence for surrogate validity, and it covers only reducible learners, I ask the authors either to add a formal equivalence analysis (e.g., confidence intervals for the difference in means with a pre-specified margin) or to soften the conclusion to 'no material difference was observed in these simulations.'
  3. [Section 4.7, Table 6] The instance-level analysis computes correlations between surrogate SHAP values and tau_hat on the same data used to fit both the CATE models and the surrogate. As a result, high values of cor(phi_j, tau_hat) may reflect in-sample overfitting rather than a genuine local biomarker signal, and the P_win metric inherits this problem. Reporting cross-validated or out-of-sample SHAP correlations, or at least the stability of the rankings across refits, would make the local-inference claims more convincing.
minor comments (6)
  1. [Section 2.3.2 and Section 6] The Discussion states that X-learner fits a single final model allowing direct SHAP, while Section 2.3.2 classifies X-learner as an irreducible meta-learner; these descriptions should be reconciled.
  2. [Section 3.3] The sentence 'For small numbers of covariates p, i.e. > 8−15' is self-contradictory; likely '<8-15' or 'up to 8-15' was intended.
  3. [Section 4.3, Table 3 and Figure 6] The text in Section 4.3 refers to Table 3 and Figure 5 for the main metrics, then separately to Figure 6 for the correlation analysis; the cross-references should be tightened so readers do not confuse the two displays.
  4. [Section 4.4] The text refers to 'the five truly prognostic covariates,' but k1(x) contains seven distinct prognostic covariates (x1, x2, and x5-x9, with x2 categorical); the count should be corrected.
  5. [Appendix] The computational-cost table is labeled 'Table 1,' duplicating the notation of the main-text Table 1; renumber it (e.g., Table A1).
  6. [Section 5.1] The sentence 'This closely replicates the ranking from earlier investigations, where VIPs from Causal Forest preferred Age over Pansspos' would be clearer if 'earlier investigations' named the reference explicitly in that sentence.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation; surrogate SHAP benchmark is validated against external simulation truth, with only minor non-load-bearing self-citation.

full rationale

Walking the derivation chain, the central contribution is a protocol (fit an XGBoost surrogate M(2) to estimated CATE, then derive TreeSHAP values) plus a simulation benchmark against known truth. The benchmark does not use the surrogate's output to define the target: predictive status of X3 and X4 is fixed externally by the data-generating equations (9)-(12), and the metrics TOP1, NET3, and MARGIN compare SHAP-based rankings to that external ground truth. Thus no fitted parameter is renamed as a prediction, and no target quantity is defined in terms of SHAP. The only self-references are to Lipkovich et al. (ref 6) for the simulation setup and to ref 39 for prior observations about T-learner behavior; these are shared design and motivational citations, not load-bearing evidence for the paper's conclusions. The acknowledged limitation in Section 6—'the model itself can be an arbitrarily bad estimator of CATE'—and the lack of a direct fidelity check of M(2) against tau_hat for S- and X-learners in the high-dimensional benchmark are validity concerns, not circularity: the empirical rankings would still be honest measurements of the surrogate-based pipeline, even if they would not license conclusions about the underlying learner. Score 1 reflects the minor self-reference in the simulation design without any circular reduction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central benchmark conclusions depend on the calibrated simulation constants, the standard causal identification assumptions, and the unverified assumption that the XGBoost surrogate preserves the CATE estimator's importance structure. No new entities are introduced; surrogate SHAP is a procedure, not a postulated entity.

free parameters (3)
  • Simulation constants a, b, c, d in g1 and g2 = a=0.625, b=5, c=0.625, d=20
    Chosen to calibrate E[tau(X)]=0.0119 and prevalence of S_true=0.330; these constants define the simulation scenarios S2/S3 that all benchmark conclusions depend on (Section 4.1, Eqs. 11-12).
  • Propensity calibration alpha1, beta1 = -2.4, -0.2
    Calibrated to yield a 1:3 treated-to-control ratio in observational scenario S3 (Eq. 13); conclusions for S3 depend on this choice.
  • XGBoost hyperparameters = CV-selected, not reported
    All base learners and surrogates use XGBoost with extensive hyperparameter tuning over eta, gamma, max_depth, colsample, subsample, min_child_weight, and nrounds, but exact grids are not given (Sections 2.4 and 3.1).
assumptions (3)
  • domain assumption Standard causal identification assumptions hold: consistency, no unmeasured confounding, positivity.
    CATE estimation from observational data (scenario S3) requires these; in randomized scenario S2, randomization ensures exchangeability. Invoked in Section 2.1.
  • ad hoc to paper The surrogate XGBoost model faithfully represents the estimated CATE surface, and TreeSHAP values from the surrogate reflect true predictive biomarker importance.
    Strategy 3 (Section 2.4) depends on this; the paper notes SHAP explains the model, not the truth, so if the CATE model or surrogate is poor, rankings can mislead.
  • standard math TreeSHAP yields exact Shapley values for the surrogate model and these inherit the desired axioms.
    Section 2.2.4, used to justify avoiding KernelSHAP; however, axioms may not carry over to the two-stage procedure because the surrogate is not the original CATE estimator.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Overview and practical recommendations on using Shapley Values for identifying predictive biomarkers via CATE modeling." pith.science (2026). https://pith.science/paper/V7RUEHOQ

@misc{pith2026250501145,
  author       = {Pith},
  title        = {Pith review of: Overview and practical recommendations on using Shapley Values for identifying predictive biomarkers via CATE modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V7RUEHOQ}},
  note         = {Machine review of arXiv:2505.01145}
}
read the original abstract

In recent years, two parallel research trends have emerged in machine learning, yet their intersections remain largely unexplored. On one hand, there has been a significant increase in literature focused on Individual Treatment Effect (ITE) modeling, particularly targeting the Conditional Average Treatment Effect (CATE) using meta-learner techniques. These approaches often aim to identify causal effects from observational data. On the other hand, the field of Explainable Machine Learning (XML) has gained traction, with various approaches developed to explain complex models and make their predictions more interpretable. A prominent technique in this area is Shapley Additive Explanations (SHAP), which has become mainstream in data science for analyzing supervised learning models. However, there has been limited exploration of SHAP application in identifying predictive biomarkers through CATE models, a crucial aspect in pharmaceutical precision medicine. We address inherent challenges associated with the SHAP concept in multi-stage CATE strategies and introduce a surrogate estimation approach that is agnostic to the choice of CATE strategy, effectively reducing computational burdens in high-dimensional data. Using this approach, we conduct simulation benchmarking to evaluate the ability to accurately identify biomarkers using SHAP values derived from various CATE meta-learners and Causal Forest.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 59 canonical work pages

  1. [1]

    DigitalFinance 2021;3:99–148

    JacobD.CATEmeetsML.Theconditionalaveragetreatmenteffectandmachinelearning. DigitalFinance 2021;3:99–148. doi: https://doi.org/10.1007/s42521-021-00033-7

    doi:10.1007/s42521-021-00033-7 2021

  2. [2]

    ProceedingsofMachineLearning Research2017: 1–13

    GutierrezP,GérardyJY.Causalinferenceandupliftmodelling:Areviewoftheliterature. ProceedingsofMachineLearning Research2017: 1–13

    work page

  3. [3]

    Towards optimal doubly robust estimation of heterogeneous causal effects.Electronic Journal of Statistics 2023; 17(2): 3008–3049

    Kennedy EH. Towards optimal doubly robust estimation of heterogeneous causal effects.Electronic Journal of Statistics 2023; 17(2): 3008–3049

    work page 2023

  4. [5]

    Subgroup identification in clinical trials: an overview of available methods and their implementations with R.Annals of translational medicine2018; 6(7)

    Zhang Z, Seibold H, Vettore MV, Song WJ, François V. Subgroup identification in clinical trials: an overview of available methods and their implementations with R.Annals of translational medicine2018; 6(7)

    work page

  5. [6]

    Modern approaches for evaluating treatment effect heterogeneity from clinical trials and observational data.Statistics In Medicine2024; 43(22): 4388-4436

    Lipkovich I, Svensson D, Ratitch B, Dmitrienko A. Modern approaches for evaluating treatment effect heterogeneity from clinical trials and observational data.Statistics In Medicine2024; 43(22): 4388-4436. doi: DOI: 10.1002/sim.10167

    doi:10.1002/sim.10167

  6. [7]

    Biometrics2017; 73: 1199-1209

    ChenS,TianL,CaiT,YuM.Ageneralstatisticalframeworkforsubgroupidentificationandcomparativetreatmentscoring. Biometrics2017; 73: 1199-1209. doi: DOI: 10.1111/biom.12676

    doi:10.1111/biom.12676

  7. [8]

    Metalearners for estimating heterogeneous treatment effects using machine learning

    Künzel SR, Sekhona JS, Bickela PJ, Yua B. Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the National Academy of Sciences (PNAS)2019; 116(10). doi: 10.1093/ectj/utaa014

    doi:10.1093/ectj/utaa014 2019

  8. [9]

    Cross-fitting and averaging for machine learning estimation of heterogeneous treatment effects

    Jacob D. Cross-fitting and averaging for machine learning estimation of heterogeneous treatment effects. arXiv:2007.02852v22020

    arXiv 2007

Show all 75 references
  1. [10]

    Subgroup identification using the personalized package.Journal of Statistical Software2021; 98(5): 1-–60

    Huling JD, Yu M. Subgroup identification using the personalized package.Journal of Statistical Software2021; 98(5): 1-–60

  2. [11]

    ObservationalStudies 2016;5(2):37–51

    AtheyS,WagerS.Estimatingtreatmenteffectswithcausalforests:anapplication. ObservationalStudies 2016;5(2):37–51

    2016

  3. [12]

    Some methods for heterogeneous treatment effect estimation in high dimensions.Statistics in Medicine2018; 37(11): 1767–1787

    Powers S, Qian J, Jung K, et al. Some methods for heterogeneous treatment effect estimation in high dimensions.Statistics in Medicine2018; 37(11): 1767–1787

  4. [13]

    CRAN brf package: Causal Inference for a Binary Treatment and Continuous Outcome using Bayesian Causal Forests

    Hahn P, Murray J, Carvalho C. CRAN brf package: Causal Inference for a Binary Treatment and Continuous Outcome using Bayesian Causal Forests. 2022. doi: https://CRAN.R-project.org/package=bcf

    2022

  5. [14]

    Predictive biomarker identification for biopharmaceutical development.Statistics in Biopharmaceutical Research2021; 13(2): 239–247

    Xin H, Hesen L, Yihua G, Ivan CS. Predictive biomarker identification for biopharmaceutical development.Statistics in Biopharmaceutical Research2021; 13(2): 239–247. doi: DOI: 10.1080/19466315.2020.1819404

    2020

  6. [15]

    Gottlow M, Svensson D, Lipkovich I, et al. Application of structured statistical analyses to identify a biomarker predictive of enhanced tralokinumab efficacy in phase III clinical trials for severe, uncontrolled asthma.BMC Pulmonary Medicine, Springer Science and Business Med...

    doi

  7. [16]

    DataMining Knowl Discov2019; 9(5)

    LohWY,CaoL,ZhouP.Subgroupidentificationforprecisionmedicine:Acomparativereviewof13methods. DataMining Knowl Discov2019; 9(5). doi: https://doi.org/10.1002/widm.1326

    doi

  8. [17]

    35thConferenceonNeuralInformationProcessingSystems(NeurIPS), Track on Datasets and Benchmarks2021

    CurthA,SvenssonD,WeatherallJ,Schaarv.dM.ReallyDoingGreatatEstimatingCATE?ACriticalLookatMLBench- markingPracticesinTreatmentEffectEstimation.. 35thConferenceonNeuralInformationProcessingSystems(NeurIPS), Track on Datasets and Benchmarks2021

  9. [18]

    WATCH: A Workflow to Assess Treatment Effect Heterogeneity in Drug Development for Clinical Trial Sponsors.Pharmaceutical Statistics2025; 24(2): e2463

    Sechidis K, Sun S, Chen Y, et al. WATCH: A Workflow to Assess Treatment Effect Heterogeneity in Drug Development for Clinical Trial Sponsors.Pharmaceutical Statistics2025; 24(2): e2463. doi: https://doi.org/10.1002/pst.2463

    doi

  10. [19]

    Distinguishing prognostic and predictive biomarkers: an information theoretic approach.Bioinformatics 2018; 34(23)

    Sechidis K, Metcalfe P, Svensson D, Weatherall J, Brown G. Distinguishing prognostic and predictive biomarkers: an information theoretic approach.Bioinformatics 2018; 34(23). doi: doi: 10.1093/bioinformatics/bty515 AUTHOR ONEET AL 33

    2018 doi

  11. [20]

    Springer New York, NY

    Hastie T, Tibshirani R, Friedman J.The Elements of Statistical Learning. Springer New York, NY. 2 ed. 2009

    2009

  12. [21]

    A unified approach to interpreting model predictions..Advances in Neural Information Processing Systems2017; 30

    Lundberg S, Lee SI. A unified approach to interpreting model predictions..Advances in Neural Information Processing Systems2017; 30

  13. [22]

    Applied Causal Inference Powered by ML and AI.arxiv (book manuscript)2024

    Chernozhukov V, Hansen C, Kallus N, Spindler M, Syrgkanis V. Applied Causal Inference Powered by ML and AI.arxiv (book manuscript)2024. doi: https://arxiv.org/abs/2403.02467

    2024 arXiv

  14. [23]

    In: International Committee on Computational Linguistics

    MoscaE,SzigetiF,TragianniS,GallagherD,GrohG.SHAP-BasedExplanationMethods:AReviewforNLPInterpretabil- ity. In: International Committee on Computational Linguistics. ; 2022; Gyeongju, Republic of Korea: 4593–4603

    2022

  15. [24]

    Causal Shapley Values: Exploiting Causal Knowledge to Explain Individual Predictions of Complex Models

    Heskes T, Sijben E, Bucur IG, Claassen T. Causal Shapley Values: Exploiting Causal Knowledge to Explain Individual Predictions of Complex Models. arXiv preprint arXiv:2011.01625; 2020

    2011 arXiv

  16. [25]

    Variable importance measures for heterogeneous causal effects.arXiv preprint arXiv:2204.060302022

    Hines O, Diaz-Ordaz K, Vansteelandt S. Variable importance measures for heterogeneous causal effects.arXiv preprint arXiv:2204.060302022

  17. [26]

    Subgroup identification from randomized clinical trial data.Statistics in Medicine2011; 30(24): 2867–2880

    Foster JC, Taylor JM, Ruberg SJ. Subgroup identification from randomized clinical trial data.Statistics in Medicine2011; 30(24): 2867–2880

  18. [27]

    Man M, Nguyen TS, Battioui C, Mi G.Predictive Subgroup/Biomarker Identification and Machine Learning Methods: 1–22; Cham: Springer International Publishing . 2019

    2019

  19. [28]

    Bagging predictors.Machine learning1996; 24(2): 123–140

    Breiman L. Bagging predictors.Machine learning1996; 24(2): 123–140

  20. [29]

    TheAnnalsofAppliedStatistics 2010; 4(1): 266 – 298

    ChipmanHA,GeorgeEI,McCullochRE.BART:Bayesianadditiveregressiontrees. TheAnnalsofAppliedStatistics 2010; 4(1): 266 – 298. doi: 10.1214/09-AOAS285

    2010 doi

  21. [30]

    Classification and regression trees.Wiley interdisciplinary reviews: data mining and knowledge discovery2011; 1(1): 14–23

    Loh WY. Classification and regression trees.Wiley interdisciplinary reviews: data mining and knowledge discovery2011; 1(1): 14–23

  22. [31]

    Estimating causal effects of treatments in randomized and nonrandomized studies.Journal of Educational Psychology1974; 66(5): 688–701

    Rubin DB. Estimating causal effects of treatments in randomized and nonrandomized studies.Journal of Educational Psychology1974; 66(5): 688–701

  23. [32]

    Generic machine learning inference on heterogeneous treatment effects in randomized experiments, with an application to immunization in India.NBER Working Paper2018(No

    Chernozhukov V, Demirer M, Duflo E, Fernandez-Val I. Generic machine learning inference on heterogeneous treatment effects in randomized experiments, with an application to immunization in India.NBER Working Paper2018(No. 24678)

  24. [33]

    Lipkovich I, Dmitrienko A, Sr. RBD. Tutorial in biostatistics: data-driven subgroup identification and analysis in clinical trials. Statistics in Medicine2017; 36: 136–196. doi: 10.1002/sim.7064

    doi

  25. [34]

    WIREsData Mining and Knowledge Discovery2019; 9(5): e1326

    LohWY,CaoL,ZhouP.Subgroupidentificationforprecisionmedicine:Acomparativereviewof13methods. WIREsData Mining and Knowledge Discovery2019; 9(5): e1326

  26. [35]

    Validating Causal Inference Models via Influence Functions

    Alaa A, Van Der Schaar M. Validating Causal Inference Models via Influence Functions. In: Chaudhuri K, Salakhutdinov R., eds.Proceedings of the 36th International Conference on Machine Learning. 97 ofProceedings of Machine Learning Research. PMLR. ; 2019: 191–201

    2019

  27. [36]

    StatisticsinBiopharmaceuticalResearch 2015; 7(3): 214–229

    RubergSJ,ShenL.Personalizedmedicine:fourperspectivesoftailoredmedicine. StatisticsinBiopharmaceuticalResearch 2015; 7(3): 214–229

    2015

  28. [37]

    Generalized random forests.The Annals of Statistics2019; 47(2): 1148–1178

    Athey S, Tibshirani J, Wager S. Generalized random forests.The Annals of Statistics2019; 47(2): 1148–1178

  29. [38]

    doi: https://doi.org/10.1214/19-BA1195

    HahnPR,MurrayJS,CarvalhoCM.Bayesianregressiontreemodelsforcausalinference:regularization,confounding,and heterogeneous effects (with discussion).Bayesian Anal.2020; 15(3): 965–1056. doi: https://doi.org/10.1214/19-BA1195

    2020 doi

  30. [39]

    On discovering treatment-effect modifiers using Virtual Twins and Causal Forest ML in the presence of prognostic biomarkers

    Hermansson E, Svensson D. On discovering treatment-effect modifiers using Virtual Twins and Causal Forest ML in the presence of prognostic biomarkers. In: Gervasi O, Murgante B, Misra S, et al., eds.Computational Science and Its Applications – ICCSA 2021. Springer Internationa...

    2021

  31. [40]

    rlearner: Quasi-Oracle Estimation of Heterogeneous Treatment Effects

    Nie X, Schuler A, Wager S. rlearner: Quasi-Oracle Estimation of Heterogeneous Treatment Effects. https://github.com/ xnie/rlearner; 2023

    2023

  32. [41]

    Root- N-Consistent Semiparametric Regression.Econometrica1988; 56(4): 931–54

    Robinson P. Root- N-Consistent Semiparametric Regression.Econometrica1988; 56(4): 931–54

  33. [42]

    Recursive partitioning for heterogeneous causal effects.Proceedings of the National Academy of Sciences 2016; 113(27): 7353–7360

    Athey S, Imbens G. Recursive partitioning for heterogeneous causal effects.Proceedings of the National Academy of Sciences 2016; 113(27): 7353–7360

    2016

  34. [43]

    A simple method for estimating interactions between a treatment and a large number of covariates.Journal of the American Statistical Association2014; 109(508): 1517–1532

    Tian L, Alizadeh AA, Gentles AJ, Tibshirani R. A simple method for estimating interactions between a treatment and a large number of covariates.Journal of the American Statistical Association2014; 109(508): 1517–1532

  35. [44]

    M.Interpretable Machine Learning: A Guide for Making Black Box Models Explainable

    C. M.Interpretable Machine Learning: A Guide for Making Black Box Models Explainable. 2019

    2019

  36. [45]

    Molnar C, Casalicchio G, Bischl B.Interpretable Machine Learning – A Brief History, State-of-the-Art and Challenges: 417–431; Springer International Publishing . 2020

    2020

  37. [46]

    Whyshoulditrustyou?

    RibeiroMT,SinghS,GuestrinC."Whyshoulditrustyou?"Explainingthepredictionsofanyclassifier.In:ACMSIGKDD. ; 2016: 1135–1144

    2016

  38. [47]

    Deep inside convolutional networks: visualising image classification models and saliency maps

    Simonyan K, Vedaldi A, Zisserman A. Deep inside convolutional networks: visualising image classification models and saliency maps. In: ICLR. ; 2014

    2014

  39. [48]

    2017: 3145–3153

    ShrikumarA,GreensideP,KundajeA.Learningimportantfeaturesthroughpropagatingactivationdifferences.In:PMLR. ; 2017: 3145–3153

    2017

  40. [49]

    Grad-cam: Visual explanations from deep networks via gradient-based localization

    Selvaraju RR, Cogswell M, Das A, Vedantam R, Parikh D, Batra D. Grad-cam: Visual explanations from deep networks via gradient-based localization. In: IEEE. ; 2017: 618–626

    2017

  41. [50]

    Smoothgrad: removing noise by adding noise.arXiv preprint arXiv:1706.038252017

    Smilkov D, Thorat N, Kim B, Viégas F, Wattenberg M. Smoothgrad: removing noise by adding noise.arXiv preprint arXiv:1706.038252017

    arXiv

  42. [51]

    A Guide for Making Black Box Models Explainable

    Molnar C.Interpretable Machine Learning. A Guide for Making Black Box Models Explainable. 2 ed. 2024

    2024

  43. [52]

    A Value for n-Person Games: 307–318; Princeton: Princeton University Press

    Shapley LS.17. A Value for n-Person Games: 307–318; Princeton: Princeton University Press . 1953

    1953

  44. [53]

    The many Shapley values for model explanation

    Sundararajan M, Najmi A. The many Shapley values for model explanation. arXiv preprint arXiv:1908.08474; 2020

    1908 arXiv

  45. [55]

    Greedy function approximation: a gradient boosting machine.Annals of statistics2001: 1189–1232

    Friedman JH. Greedy function approximation: a gradient boosting machine.Annals of statistics2001: 1189–1232

  46. [56]

    Nature machine intelligence2020; 2(1): 56–67

    LundbergSM,ErionG,ChenH,etal.FromlocalexplanationstoglobalunderstandingwithexplainableAIfortrees. Nature machine intelligence2020; 2(1): 56–67

  47. [57]

    Improving the Sampling Strategy in KernelSHAP.arXiv preprint arXiv:2410.048832024

    Olsen LHB, Jullum M. Improving the Sampling Strategy in KernelSHAP.arXiv preprint arXiv:2410.048832024

  48. [58]

    Algorithms to estimate Shapley value feature attributions.Nature Machine Intelligence2023; 5(6): 590–601

    Chen H, Covert IC, Lundberg SM, Lee SI. Algorithms to estimate Shapley value feature attributions.Nature Machine Intelligence2023; 5(6): 590–601

  49. [59]

    kernelshap: Kernel SHAP. 2024. R package version 0.7.0

    2024

  50. [60]

    CRAN; CRAN: 2021

    Tibshirani J, Athey S, Sverdrup E, Wager S.grf: Generalized Random Forests. CRAN; CRAN: 2021. R package version 2.0.2

    2021

  51. [61]

    shapr: An R-package for explaining machine learning models with dependence-aware Shapley values.Journal of Open Source Software2019; 5(46): 2027

    Sellereite N, Jullum M. shapr: An R-package for explaining machine learning models with dependence-aware Shapley values.Journal of Open Source Software2019; 5(46): 2027. doi: 10.21105/joss.02027

    2027 doi

  52. [62]

    SuperLearner: Super Learner Prediction. 2024. R package version 2.0-29

    2024

  53. [63]

    Model-agnostic interpretability with shapley values

    Messalas A, Kanellopoulos Y, Makris C. Model-agnostic interpretability with shapley values. In: IEEE. ; 2019: 1–7. AUTHOR ONEET AL 35

    2019

  54. [64]

    SHAP-Based Explanation Methods: A Review for NLP Inter- pretability

    Mosca E, Szigeti F, Tragianni S, Gallagher D, Groh G. SHAP-Based Explanation Methods: A Review for NLP Inter- pretability. Proceedings of the 29th International Conference on Computational Linguistics2022: 4593–4603. doi: https://doi.org/10.1016/j.artint.2021.103502

  55. [65]

    Tabular data: Deep learning is not all you need.Information Fusion2022; 81: 84-90

    Shwartz-Ziv R, Armon A. Tabular data: Deep learning is not all you need.Information Fusion2022; 81: 84-90. doi: https://doi.org/10.1016/j.inffus.2021.11.011

    2021 doi

  56. [66]

    Why do tree-based models still outperform deep learning on typical tabular data?

    Grinsztajn L, Oyallon E, Varoquaux G. Why do tree-based models still outperform deep learning on typical tabular data?. In: Neural Information Processing Systems. ; 2022

    2022

  57. [67]

    CRAN; CRAN: 2023

    Chen T, He T, Benesty M, et al.xgboost: Extreme Gradient Boosting. CRAN; CRAN: 2023. R package version 1.7.5.1

    2023

  58. [68]

    Classification and Regression by randomForest.R News2002; 2(3): 18-22

    Liaw A, Wiener M. Classification and Regression by randomForest.R News2002; 2(3): 18-22

  59. [69]

    XGBoost: A Scalable Tree Boosting System

    Chen T, Guestrin C. XGBoost: A Scalable Tree Boosting System. In: KDD ’16. Association for Computing Machinery. ACM; 2016

    2016

  60. [70]

    Overview of modern approaches for identifying and evaluating heterogeneous treatment effects from clinical data.Clinical Trials2023; 20(4)

    Lipkovich I, Svensson D, Ratitch B, Dmitrienko A. Overview of modern approaches for identifying and evaluating heterogeneous treatment effects from clinical data.Clinical Trials2023; 20(4). doi: DOI: 10.1177/17407745231174544

    doi

  61. [71]

    Random Forests of Interaction Trees for Estimating Individualized Treatment Effects in Randomized Trials.Statistics in Medicine2017; 37

    Su X, Peña A, Liu L, Levine R. Random Forests of Interaction Trees for Estimating Individualized Treatment Effects in Randomized Trials.Statistics in Medicine2017; 37. doi: 10.1002/sim.7660

    doi

  62. [73]

    Acupuncture for chronic headache in primary care: large, pragmatic, randomised trial

    Vickers A, Rees RW, Zollman CE, et al. Acupuncture for chronic headache in primary care: large, pragmatic, randomised trial. BMJ2004. doi: doi:10.1136/bmj.38029.421863.EB

  63. [74]

    Model-based recursive partitioning for subgroup analyses.The International Journal of Biostatistics2016; 12(1): 45–63

    Seibold H, Zeileis A, Hothorn T. Model-based recursive partitioning for subgroup analyses.The International Journal of Biostatistics2016; 12(1): 45–63

  64. [75]

    Distilling heterogeneous treatment effects: Stable subgroup estimation in causal inference.arxiv 2025

    Huang M, Tang TM, Kennedy AM. Distilling heterogeneous treatment effects: Stable subgroup estimation in causal inference.arxiv 2025

    2025

  65. [76]

    Experimental evaluation of individualized treatment rules.Journal of the American Statistical Association 2021; 0(0): 1-15

    Imai K, Li ML. Experimental evaluation of individualized treatment rules.Journal of the American Statistical Association 2021; 0(0): 1-15

    2021

  66. [77]

    LLpowershap: logistic loss-based automated Shapley values feature selection method.BMC Medical Research Methodology2024; 24(1): 247

    Madakkatel I, Hyppönen E. LLpowershap: logistic loss-based automated Shapley values feature selection method.BMC Medical Research Methodology2024; 24(1): 247

  67. [78]

    In: Neural Information Processing Systems

    MuschalikM,BanieckiH,FumagalliF,KolpaczkiP,HammerB,HüllermeierE.shapiq:ShapleyInteractionsforMachine Learning. In: Neural Information Processing Systems. ; 2024

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.