Pith. sign in

REVIEW 4 major objections 5 minor 47 references

M-learner:A Flexible And Powerful Framework To Study Heterogeneous Treatment Effect In Mediation Model

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The M-learner is the first framework to estimate per-person mediated treatment effects and cluster patients into subtypes with distinct indirect treatment effects.

desk verdict A useful exploratory pipeline for subgroup discovery in mediation settings, but its claimed Type I error control rests on an unproven chi-squared assumption for a data-adaptively selected tree. read the letter →

arxiv 2505.17917 v3 pith:I3EK2BNS submitted 2025-05-23 stat.ML cs.LGstat.ME

Xingyu Li , Qing Liu , Tony Jiang , Hong Amy Xia , Brian P. Hobbs , Peng Wei This is my paper · ORCID
classification stat.MLcs.LGstat.ME MSC 62H3062G08
keywords heterogeneoustreatmenteffectsmediationanalysissubgroupidentificationconditionalindirecteffectmeta-learnerscausalinferenceTypeIerrorcontrolt-SNEprojection
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 proposes the M-learner, an algorithm that estimates how much of a treatment's effect runs through a mediator variable at the level of individual patients, then groups patients into covariate-defined subtypes that share the same indirect treatment effect. The claim matters because clinical trials often track mediators as early surrogate signals of treatment success, yet existing heterogeneous-treatment-effect methods ignore the mediation structure, so a treatment that works through the mediator in only a subset of patients would be averaged away. The M-learner is, to the authors' knowledge, the first framework that both tests for the presence of mediated heterogeneity and identifies which subgroups carry it, while holding the false-positive rate at a pre-specified level. In simulations it recovers the true heterogeneity-driving covariates in nearly all simple-setting runs, keeps Type I error at or below the 10% calibration level under null settings, and on the JOBS II trial it isolates a large subtype in which the mediator is significant even though the overall mediation test is not.

What carries the argument

The load-bearing object is the per-unit treatment-effect estimate turned into a geometry. The distance $dis(i,j) = (\hat{\tau}_i - \hat{\tau}_j)^2$ between two units' effect estimates is assembled into a full $n \times n$ matrix, and this matrix — not the raw covariates — carries the signal of heterogeneity. t-SNE, a projection method that places similar points close together in a low-dimensional map, projects the matrix into two Euclidean dimensions; K-means partitions the projected points; and a decision tree re-expresses each cluster as a rule on covariates whose leaves are the 'subtypes.' The calibration step compares the regressions $M = \beta_1 \mathrm{leaf} + \beta_2 W$ and $M = \beta_3 \mathrm{leaf} + \beta_4 W + \beta_5 \mathrm{leaf} \cdot W$; the statistic $2(\log L_1 - \log L_0)$ is assumed chi-squared with degrees of freedom equal to the number of leaves minus one, and the clustering with the smallest $p_{\text{leaf}}$ is retained. When there is no mediator, $Y$ replaces $M$ in that comparison, so the same machinery covers subgroups defined by the total treatment effect.

What would settle it

Simulate many datasets under the paper's null scenarios (Simple-Null1 and Simple-Null2), run the full M-learner pipeline, and replace the chi-squared $p_{\text{leaf}}$ with a permutation null obtained by randomly reassigning units to the tree's leaves while keeping treatment labels fixed; if the rejection rate at the nominal 10% threshold visibly exceeds 10% over thousands of runs, the claimed Type I error control is an artifact of double use of the data.

Watch

Extended reading notes

Core claim

The paper's central claim is that treatment-effect heterogeneity inside a mediation model can be detected, localized, and tested by a four-step pipeline: estimate each unit's conditional average indirect or total treatment effect (CAITE or CATTE), form an $n \times n$ distance matrix from squared pairwise differences $(\hat{\tau}_i - \hat{\tau}_j)^2$ of those estimates, project the matrix to two dimensions with t-SNE and cluster with K-means, and translate each clustering into a decision tree whose leaves are the interpretable 'subtypes,' selecting the tree with the smallest likelihood-ratio p-value $p_{\text{leaf}}$. The indirect-effect estimate is $\hat{\tau}^{\mathrm{ITE}}(x) = \hat{g}_1^Y(x, \hat{g}_1^M(x)) - \hat{g}_1^Y(x, \hat{g}_0^M(x))$: predictions of the mediator under treatment and control are plugged into the treated-outcome regression. Across 100 simulations per scenario, the framework reports recovery of the true heterogeneity covariates in nearly all simple-setting runs, Type I error at or below the 10% calibration level under null scenarios, and identified subtype regions that run somewhat smaller than the ground-truth regions while tracking their mediation proportions within one standard deviation. On the JOBS II trial, the method finds subtypes of unemployed workers in which the mediator job-search self-efficacy carries a significant indirect effect (for example, $p = 0.010$ in a 344-person subtype) even though the overall mediation analysis reports a non-significant indirect effect ($p = 0.21$), which the paper presents as evidence that global mediation tests can hide localized mediated effects.

Load-bearing premise

The p-value that decides which subtype grouping is real is computed assuming a chi-squared distribution, even though the same data first chose the clusters and tree leaves and then were used to test them; if that assumption is wrong, the claimed 10% error control and the reported p-values are not valid.

Editorial extensions

If this is right

  • Trials that track a surrogate mediator can ask not just whether a mediated effect exists but in which covariate-defined patients the mediator carries it; the paper argues this enables monitoring mediators within subtypes and stopping ineffective interventions early for particular groups.
  • The estimation step is independent of model architecture, so any supervised learner plugs in — the paper demonstrates random forests and XGBoost and names neural networks — letting users match the learner to the data's structure.
  • The authors state that individual treatment responsiveness can be inferred from the mediator without observing the final outcome $Y$, which they connect to adaptive interventions in technology companies, personalized medicine, and government trials.
  • In the absence of a mediator the framework reduces to a T-learner subgroup analysis for the total treatment effect, so the same pipeline serves both mediated and unmediated studies.
  • On the JOBS II application, the mediator's indirect effect moves from non-significant overall ($p = 0.21$) to significant ($p = 0.010$) within the largest subtype, a pattern the paper offers as evidence that global mediation analyses can mask localized mediated effects.

Reading between the lines

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

  • Because the same data select the K-means clusters and the decision-tree leaves that feed the chi-squared calibration, the nominal $p_{\text{leaf}}$ may overstate significance; a permutation null that randomly reassigns units across leaves, or a split-sample calibration, would show how much of the reported error control survives the data-adaptive selection.
  • The distance-matrix-plus-projection recipe is generic: the same geometry could be built from conditional direct effects, quantile treatment effects, or time-to-event contrasts, turning any per-unit causal estimate into a subgroup-discovery pipeline.
  • If the JOBS II pattern generalizes, power analyses that assume a single global mediation effect are mis-specified, and mediator-based subgroup discovery should come before go/no-go decisions in confirmatory trials.
  • t-SNE is stochastic, so a single run's subgroups may not be reproducible; a stability audit that repeats the pipeline across random seeds and measures how often the same leaf rule is selected would tell practitioners how much to trust one run.
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

4 major / 5 minor

Summary. The paper proposes the 'M-learner', a four-step pipeline for estimating heterogeneous treatment effects in mediation settings and for identifying subgroups with distinct indirect or total treatment effects. The steps are: (1) estimate per-unit conditional average total or indirect treatment effects (CATTE/CAITE) with a base learner; (2) compute a pairwise Euclidean distance matrix from the estimated effects; (3) project the distance matrix via t-SNE and cluster with K-means; (4) fit decision trees to the clusters, select the tree with the smallest p-value p_leaf computed from a likelihood-ratio test, and calibrate a threshold under a null scenario to control the Type I error rate. The paper reports simulations in mediated and unmediated settings, sensitivity analyses, comparisons with K-means and other learners, and an application to the JOBS II dataset.

Significance. If the statistical guarantees were valid, the framework would be a useful contribution: it is, to my knowledge, one of the first explicit frameworks for detecting and localizing treatment-effect heterogeneity transmitted through a mediator, it is agnostic to the base learner, and it ships with code and a real-data demonstration. The simulation study is extensive, covering several heterogeneity structures, sample sizes, noise levels, projection methods, and clustering baselines. The central inferential claim, however, is the control of the Type I error rate at a pre-specified level, and that claim rests on a distributional assumption for p_leaf that is asserted rather than derived and that is used after data-driven selection of clusters and tree partitions. The paper also contains an explicit admission in Section 5 that unbiasedness holds only under linearity of Y in M, but no proof or regularity conditions are given. These issues are load-bearing for the paper's main advertised contributions, so the manuscript cannot be accepted in its current form.

major comments (4)
  1. [Section 2.2, Eqs. (2)-(3)] The claim that 2(log L1 - log L0) follows a chi-squared distribution with degrees of freedom equal to the number of leaves minus one is not justified for the actual pipeline. The leaves are not fixed in advance: the same data produce the estimated treatment effects, the distance matrix, the t-SNE projection, the K-means solutions for K = 2..5, and the decision-tree partitions, and then the tree with the smallest p_leaf is selected. Under the null of no heterogeneity, the reported statistic is therefore a minimum of several correlated likelihood-ratio statistics with data-dependent leaf definitions, so its null distribution need not be chi-squared. Since p_leaf is used both for final tree selection and for the claimed 10% Type I error calibration in Tables 1-2, this is a load-bearing step. The authors should either derive a valid null distribution for the full selection pipeline or replace the chi-squared calibration with a resampling or sample-splitting procedure that is valid under data-driven selection.
  2. [Section 3, Tables 1 and 2; Appendix A.4] The calibration protocol sets a threshold under one Null scenario simulated from the same generative family and then applies that threshold to other scenarios. This only shows that the empirical false-positive count is near the calibration level in that particular null setting; it does not establish 'Type I error control at a pre-specified level' in the sense usually claimed for a testing procedure. The null hypothesis being tested is not formally defined, no error-rate guarantee is stated, and no validation is provided under different null-generating mechanisms (e.g., different covariate distributions, different mediator strengths, or different noise levels). The paper should state the null hypothesis explicitly and either prove a finite-sample or asymptotic error-rate bound or reframe the procedure as empirical calibration, with the caveat that the reported rates apply only to the simulated null family used for calibration.
  3. [Section 2.2 and Section 5 (unbiasedness claim)] The statement in Section 5 that 'Our method is unbiased when the Y has linear relationship with mediator M' is not established by any derivation or set of regularity conditions. The estimator tau_hat^ITE(x) = g_hat_1^Y(x, g_hat_1^M(x)) - g_hat_1^Y(x, g_hat_0^M(x)) is a plug-in estimator, and finite-sample bias from estimating g_0^M, g_1^M and g_1^Y is not analyzed. Moreover, the complex scenarios in Appendix A.10.2 use a logistic outcome model, where the linearity condition fails, so the unbiasedness claim cannot cover the simulation results that are presented as supporting the method. The claim should be restated as a consistency result under appropriate regularity conditions, or removed and replaced with an explicit bias analysis.
  4. [Appendix A.10.2 and Table 3] The comparison in Table 3 selects, among the final subtype regions, the region exhibiting the most statistically significant mediation effect before recording its sample size and mediation proportion. This best-region selection will bias the reported estimates and can inflate agreement with the ground-truth heterogeneous region. The evaluation therefore does not provide an unbiased picture of the regions identified by the M-learner. A pre-specified rule for choosing the region to report, or reporting the full set of discovered regions with a correction for selection, is needed before the subgroup-identification claims in Table 3 can be accepted.
minor comments (5)
  1. [Section 2.2] The description of g_hat_0^M says 'using the treated observations {(Xi,Yi)}Wi=0', which appears to be a typo: the control mediator function should be estimated from control observations, and the response should be M, not Y.
  2. [Abstract and Introduction] There are several incomplete sentences and typos, e.g., 'First, we compute individual-level conditional average indirect/total treatment effect' lacks a period, and 'This is the first method capable...' begins with an unnecessary capital 'This' after a period. The paper would benefit from a careful proofread.
  3. [Section 5] The sentence 'Importantly, our method permits the inference of individual treatment responsiveness without requiring observation of the final outcome variable Y' is potentially misleading, since the CAITE estimator in Section 2.2 uses Y to fit g_1^Y. Please clarify whether the intended claim concerns prediction for new units rather than estimation from the training data.
  4. [Appendix A.6.2 and A.7] The t-SNE implementation parameters (perplexity, learning rate, number of iterations, initialization) are not reported, so the sensitivity of the pipeline to these choices cannot be assessed from the paper; a robustness table or a fixed default setting with justification would be helpful.
  5. [Appendix A.10.1 and A.10.2] Several equations and scenario references in the appendix are broken or incomplete; for example, the scenario definitions referred to in the visualization text are not all explicitly restated, and some displayed formulas in A.10.2 are split awkwardly across lines.

Circularity Check

1 steps flagged · score 6.0 of 10

Claimed Type I error control is partly by construction: the decision threshold is fit to the null scenario, so the null rows in Tables 1–2 restate the calibration target rather than independently validating it; the chi-square claim for the data-dependent pleaf is an unproven assumption, not a derived guarantee.

  1. fitted input called prediction [Section 3.1 (calibration description) and Tables 1–2; see also Section 3.2 for Simple-Null2]
    "Specifically, we determine a threshold under the Null scenario such that the Type I error rate is controlled at 10%, and apply this threshold to assess the validity of subtype groupings in other scenarios. ... In contrast to the unmediated setting, we calibrate the threshold based on Scenario Simple-Null2 by controlling the Type I error at 10%, and then apply this threshold to assess the validity of subtype identification across other scenarios."

    The threshold is fitted to the Null (or Simple-Null2) simulation so that the empirical Type I error is at or below 10%. The Null rows of Tables 1 and 2 (e.g., RF Null: X1=4, X2=4, X1&X2=2; XGB Simple-Null2: 2,1,0) are then reported as evidence that the method 'maintains a low Type I error rate.' But this is the calibration objective restated: any threshold chosen to satisfy the 10% constraint will, by construction, produce such counts on the same Null distribution. The paper's general claim to 'effectively control the Type I error rate' is therefore not an independent property of the M-learner; in the calibration scenario it is an enforced fitting constraint.

full rationale

The M-learner's core estimation and subgroup pipeline are not circular: CAITE is defined as a plug-in contrast \g^Y_1(x,\g^M_1(x)) − \g^Y_1(x,\g^M_0(x)) from fitted response and mediator functions, and the distance-matrix → t-SNE → K-means → decision-tree sequence is a stated algorithm, not a disguised identity. No load-bearing self-citation or imported uniqueness theorem appears; the phrase 'we use M-learner proposed in our paper' is only a label. The one constructed equivalence is in the calibration step: the threshold is chosen on the Null/Simple-Null2 simulations to make the Type I error 10%, so the corresponding rows in Tables 1–2 are the fitting target, not an independent prediction. This is a partial circularity in the paper's central 'controls Type I error' claim. Separately, the chi-square distribution asserted for pleaf after data-dependent leaf selection (Section 2.2) is a correctness risk: the statistic is a minimum over correlated, data-dependent likelihood-ratio tests, so its null distribution is not the stated chi-square with df = leaves − 1. That is an unproven distributional assumption rather than a definitional equivalence, so it does not by itself raise the circularity score beyond 6.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new scientific entities, particles, forces, or conserved quantities. Its load-bearing assumptions are the standard mediation identifiability conditions, the RCT restriction, the asserted chi-squared calibration, and the linearity condition for unbiasedness of the mediation estimator. The main free parameters are the calibrated Type I error threshold, the number-of-clusters range, and unreported t-SNE hyperparameters.

free parameters (3)
  • Type I error calibration threshold = chosen so that the Null scenario has 10 percent Type I error
    Section 3.1 states the threshold is determined under the Null scenario to control Type I error at 10%. This threshold is fitted to simulated null data and then applied to all other scenarios, so the claimed error control is by construction tied to this fitted threshold.
  • Number of clusters K range (2 to k_max) = k = floor(sqrt(d)) + 2, with experiments using up to 5 or 10
    Section 2.2 recommends k = floor(sqrt(d)) + 2 from experience, and Appendix A.6.3 shows results depend on the maximum number of clusters (K=2 gives markedly worse recovery than K=5 or K=10). This is a hand-tuned algorithmic choice that affects the central subgroup-recovery results.
  • t-SNE parameters (perplexity, learning rate, initialization) = not reported
    t-SNE is used to project the distance matrix, and its output depends strongly on hyperparameters. The paper does not report perplexity or other settings, and no sensitivity analysis covers t-SNE hyperparameters, only projection dimension.
assumptions (4)
  • domain assumption Standard mediation assumptions: no unmeasured confounding of treatment-outcome, mediator-outcome, treatment-mediator, and no mediator-outcome confounding by treatment-induced covariates
    Appendix A.1 lists assumptions (4)-(7), following VanderWeele and Vansteelandt (2009). These are standard but untestable from data.
  • domain assumption Randomized controlled trial setting with W independent of X and M
    Assumption (8) in Appendix A.1 restricts the method to RCT data, which the paper states. The JOBS II application satisfies this, but the method is not applicable to observational data without re-derivation.
  • ad hoc to paper Chi-squared distribution of 2(log L1 - log L0) with degrees of freedom equal to number of leaves minus one
    Section 2.2, equations (2)-(3), asserts this distributional result for the leaf-based test. The comparison is between a model with and without leaf-treatment interaction, but the leaves themselves are selected on the same data via clustering and decision trees, so the usual likelihood-ratio asymptotics are not justified. No proof or reference is given.
  • domain assumption Identity link / linear mediator-outcome relationship for unbiasedness
    The Discussion states the method is unbiased when Y has a linear relationship with M and has bias for nonlinear relationships. The simulation evidence in the Complex scenarios uses a nonlinear outcome but the bias is not separately reported, so the central claims rest on this linearity being mild or the bias being negligible.

how reviews work

0 comments
Cite this review

Pith. "Pith review of M-learner:A Flexible And Powerful Framework To Study Heterogeneous Treatment Effect In Mediation Model." pith.science (2026). https://pith.science/paper/I3EK2BNS

@misc{pith2026250517917,
  author       = {Pith},
  title        = {Pith review of: M-learner:A Flexible And Powerful Framework To Study Heterogeneous Treatment Effect In Mediation Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I3EK2BNS}},
  note         = {Machine review of arXiv:2505.17917}
}
read the original abstract

We propose a novel method, termed the M-learner, for estimating heterogeneous indirect and total treatment effects and identifying relevant subgroups within a mediation framework. The procedure comprises four key steps. First, we compute individual-level conditional average indirect/total treatment effect Second, we construct a distance matrix based on pairwise differences. Third, we apply tSNE to project this matrix into a low-dimensional Euclidean space, followed by K-means clustering to identify subgroup structures. Finally, we calibrate and refine the clusters using a threshold-based procedure to determine the optimal configuration. To the best of our knowledge, this is the first approach specifically designed to capture treatment effect heterogeneity in the presence of mediation. Experimental results validate the robustness and effectiveness of the proposed framework. Application to the real-world Jobs II dataset highlights the broad adaptability and potential applicability of our method.Code is available at https: //anonymous.4open.science/r/M-learner-C4BB.

Figures

Figures reproduced from arXiv: 2505.17917 by the authors.

Figure 1
Figure 1. (a)Directed Acyclic Graph between the covariates [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Empirical cumulative distribution functions (ECDF) of [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Subtype results identified by the M-learner on the Jobs II dataset. (a) It shows subtype [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 36 canonical work pages

  1. [1]

    J. D. Angrist. Treatment effect heterogeneity in theory and practice. The economic journal, 114 0 (494): 0 C52--C83, 2004

    work page 2004

  2. [2]

    Athey and G

    S. Athey and G. Imbens. Recursive partitioning for heterogeneous causal effects. Proceedings of the National Academy of Sciences, 113 0 (27): 0 7353--7360, 2016

    work page 2016

  3. [3]

    Athey and S

    S. Athey and S. Wager. Estimating treatment effects with causal forests: An application. Observational studies, 5 0 (2): 0 37--51, 2019

    work page 2019

  4. [4]

    L. Breiman. Random forests. Machine learning, 45: 0 5--32, 2001

    2001

  5. [5]

    Chen and C

    T. Chen and C. Guestrin. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, pages 785--794, 2016

    work page 2016

  6. [6]

    Cheng, R

    L. Cheng, R. Guo, and H. Liu. Causal mediation analysis with hidden confounders. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining, pages 113--122, 2022

    work page 2022

  7. [7]

    D \' az, N

    I. D \' az, N. S. Hejazi, K. E. Rudolph, and M. J. van Der Laan. Nonparametric efficient causal mediation with intermediate confounders. Biometrika, 108 0 (3): 0 627--641, 2021

    work page 2021

  8. [8]

    D \' az, N

    I. D \' az, N. Williams, and K. E. Rudolph. Efficient and flexible mediation analysis with time-varying mediators, treatments, and confounders. Journal of Causal Inference, 11 0 (1): 0 20220077, 2023

    work page 2023

Show all 47 references
  1. [9]

    J. A. Dorresteijn, F. L. Visseren, P. M. Ridker, A. M. Wassink, N. P. Paynter, E. W. Steyerberg, Y. van der Graaf, and N. R. Cook. Estimating treatment effects for individual patients based on the results of randomised clinical trials. Bmj, 343, 2011

    2011

  2. [10]

    Dwivedi, Y

    R. Dwivedi, Y. S. Tan, B. Park, M. Wei, K. Horgan, D. Madigan, and B. Yu. Stable discovery of interpretable subgroups via calibration in causal studies. International Statistical Review, 88: 0 S135--S178, 2020

    2020

  3. [11]

    J. C. Foster, J. M. Taylor, and S. J. Ruberg. Subgroup identification from randomized clinical trial data. Statistics in medicine, 30 0 (24): 0 2867--2880, 2011

    2011

  4. [12]

    L. Ge, J. Wang, C. Shi, Z. Wu, and R. Song. A reinforcement learning framework for dynamic mediation analysis. In International Conference on Machine Learning, pages 11050--11097. PMLR, 2023

    2023

  5. [13]

    L. Ge, H. Cai, R. Wan, Y. Xu, and R. Song. A review of causal decision making. arXiv preprint arXiv:2502.16156, 2025

    2025 arXiv

  6. [14]

    Gilbert, K

    B. Gilbert, K. L. Hoffman, N. Williams, K. E. Rudolph, E. J. Schenck, and I. D \' az. Identification and estimation of mediational effects of longitudinal modified treatment policies. arXiv preprint arXiv:2403.09928, 2024

    2024

  7. [15]

    Hassanpour and R

    N. Hassanpour and R. Greiner. Learning disentangled representations for counterfactual regression. In International Conference on Learning Representations, 2019

    2019

  8. [16]

    K. Imai, L. Keele, and D. Tingley. A general approach to causal mediation analysis. Psychological methods, 15 0 (4): 0 309, 2010

    2010

  9. [17]

    G. W. Imbens. Nonparametric estimation of average treatment effects under exogeneity: A review. Review of Economics and statistics, 86 0 (1): 0 4--29, 2004

    2004

  10. [18]

    Johansson, U

    F. Johansson, U. Shalit, and D. Sontag. Learning representations for counterfactual inference. In International conference on machine learning, pages 3020--3029. PMLR, 2016

    2016

  11. [19]

    E. H. Kennedy. Towards optimal doubly robust estimation of heterogeneous causal effects. Electronic Journal of Statistics, 17 0 (2): 0 3008--3049, 2023

    2023

  12. [20]

    K. Kim, J. Kim, L. Wasserman, and E. Kennedy. Hierarchical and density-based causal clustering. Advances in Neural Information Processing Systems, 37: 0 30363--30393, 2024

    2024

  13. [21]

    S. R. K \"u nzel, J. S. Sekhon, P. J. Bickel, and B. Yu. Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the national academy of sciences, 116 0 (10): 0 4156--4165, 2019

    2019

  14. [22]

    D. Lin, T. Fleming, and V. De Gruttola. Estimating the proportion of treatment effect explained by a surrogate marker. Statistics in medicine, 16 0 (13): 0 1515--1527, 1997

    1997

  15. [23]

    L. Luo, C. Shi, J. Wang, Z. Wu, and L. Li. Multivariate dynamic mediation analysis under a reinforcement learning framework. The Annals of Statistics, 53 0 (1): 0 400--425, 2025

    2025

  16. [24]

    Nie and S

    X. Nie and S. Wager. Quasi-oracle estimation of heterogeneous treatment effects. Biometrika, 108 0 (2): 0 299--319, 2021

    2021

  17. [25]

    J. Pearl. Direct and indirect effects. In Probabilistic and causal inference: the works of Judea Pearl, pages 373--392. 2022

    2022

  18. [26]

    M. L. Petersen, S. E. Sinisi, and M. J. van der Laan. Estimation of direct causal effects. Epidemiology, 17 0 (3): 0 276--284, 2006

    2006

  19. [27]

    K. J. Preacher. Advances in mediation analysis: A survey and synthesis of new developments. Annual review of psychology, 66 0 (1): 0 825--852, 2015

    2015

  20. [28]

    R. H. Price, M. Van Ryn, and A. D. Vinokur. Impact of a preventive job search intervention on the likelihood of depression among the unemployed. Journal of health and social behavior, pages 158--167, 1992

    1992

  21. [29]

    J. M. Robins and S. Greenland. Identifiability and exchangeability for direct and indirect effects. Epidemiology, 3 0 (2): 0 143--155, 1992

    1992

  22. [30]

    Shalit, F

    U. Shalit, F. D. Johansson, and D. Sontag. Estimating individual treatment effect: generalization bounds and algorithms. In International conference on machine learning, pages 3076--3085. PMLR, 2017

    2017

  23. [31]

    C. Shi, D. Blei, and V. Veitch. Adapting neural networks for the estimation of treatment effects. Advances in neural information processing systems, 32, 2019

    2019

  24. [32]

    E. J. T. Tchetgen and I. Shpitser. Semiparametric theory for causal mediation analysis: efficiency bounds, multiple robustness, and sensitivity analysis. Annals of statistics, 40 0 (3): 0 1816, 2012

    2012

  25. [33]

    E. J. T. Tchetgen and T. J. VanderWeele. Identification of natural direct effects when a confounder of the mediator is directly affected by exposure. Epidemiology, 25 0 (2): 0 282--291, 2014

    2014

  26. [34]

    Ting and A

    A. Ting and A. R. Linero. Estimating heterogeneous causal mediation effects with bayesian decision tree ensembles. Journal of the American Statistical Association, 0 (just-accepted): 0 1--20, 2025

    2025

  27. [35]

    Tingley, T

    D. Tingley, T. Yamamoto, K. Hirose, L. Keele, and K. Imai. Mediation: R package for causal mediation analysis. Journal of statistical software, 59: 0 1--38, 2014

    2014

  28. [36]

    M. J. van der Laan and M. L. Petersen. Direct effect models. The international journal of biostatistics, 4 0 (1), 2008

    2008

  29. [37]

    Van der Maaten and G

    L. Van der Maaten and G. Hinton. Visualizing data using t-sne. Journal of machine learning research, 9 0 (11), 2008

    2008

  30. [38]

    Vandenberghe, S

    S. Vandenberghe, S. Vansteelandt, and T. Loeys. Boosting the precision of mediation analyses of randomised experiments through covariate adjustment. Statistics in medicine, 36 0 (6): 0 939--957, 2017

    2017

  31. [39]

    VanderWeele

    T. VanderWeele. Explanation in causal inference: methods for mediation and interaction. Oxford University Press, 2015

    2015

  32. [40]

    VanderWeele and S

    T. VanderWeele and S. Vansteelandt. Conceptual issues concerning mediation, interventions and composition. Statistics and its Interface, 2: 0 457--468, 2009

    2009

  33. [41]

    Vansteelandt and T

    S. Vansteelandt and T. J. VanderWeele. Natural direct and indirect effects on the exposed: effect decomposition under weaker assumptions. Biometrics, 68 0 (4): 0 1019--1027, 2012

    2012

  34. [42]

    A. D. Vinokur, R. H. Price, and Y. Schul. Impact of the jobs intervention on unemployed workers varying in risk for depression. American journal of community psychology, 23 0 (1): 0 39--74, 1995

    1995

  35. [43]

    A. D. Vinokur, Y. Schul, J. Vuori, and R. H. Price. Two years after a job loss: long-term impact of the jobs program on reemployment and mental health. Journal of occupational health psychology, 5 0 (1): 0 32, 2000

    2000

  36. [44]

    Vuori and J

    J. Vuori and J. Silvonen. The benefits of a preventive job search program on re-employment and mental health at 2-year follow-up. Journal of Occupational and Organizational Psychology, 78 0 (1): 0 43--52, 2005

    2005

  37. [45]

    Wager and S

    S. Wager and S. Athey. Estimation and inference of heterogeneous treatment effects using random forests. Journal of the American Statistical Association, 113 0 (523): 0 1228--1242, 2018

    2018

  38. [46]

    Wang and R

    J. Wang and R. Song. Dynamic causal structure discovery and causal effect estimation. arXiv preprint arXiv:2501.06534, 2025

    2025 arXiv

  39. [47]

    Zhang, D

    Y. Zhang, D. Kong, and S. Yang. Towards r-learner of conditional average treatment effects with a continuous treatment: T-identification, estimation, and inference. arXiv preprint arXiv:2208.00872, 2022

    2022 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.