Pith. sign in

REVIEW 5 major objections 5 minor 61 references

Treatment Effect Estimation for Optimal Decision-Making

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Accurate CATE estimates can still yield wrong treatment decisions when the model class is misspecified, and a retargeted objective fixes the mismatch.

desk verdict The paper's core theorem has a load-bearing proof gap (unbounded M), but the proposed retargeting objective and empirical evaluation make it worth a serious revision. read the letter →

arxiv 2505.13092 v2 pith:DQH4B2PU submitted 2025-05-19 cs.LG

Dennis Frauen , Valentyn Melnychuk , Jonas Schweisthal , Mihaela van der Schaar , Stefan Feuerriegel This is my paper · ORCID
classification cs.LG
keywords conditionalaveragetreatmenteffectstwo-stagemeta-learnerspolicylearningthresholdingpoliciesdoublyrobustestimationneuralnetworkseffectheterogeneityoff-policy
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

Medical and other high-stakes decisions are often made by estimating the conditional average treatment effect (CATE) and treating everyone whose estimated CATE is positive. This paper proves that a two-stage CATE estimator (such as the DR-learner) that is optimal for estimating the CATE can still produce a strictly suboptimal thresholded policy when the true CATE falls outside the model class used in the second stage. The reason is that squared-error estimation spends effort on regions far from the decision boundary, where mistakes do not change the decision. To fix this, the paper introduces the policy-targeted CATE (PT-CATE), a retargeted estimand that blends CATE accuracy with policy value, and a neural procedure with an adaptively smoothed decision indicator. If correct, the work gives practitioners a principled way to keep interpretable CATE-based decisions while approaching the decision quality of direct policy optimization.

What carries the argument

The load-bearing object is the $\gamma$-policy-targeted CATE ($\gamma$-PT-CATE), defined as the minimizer of $L_\gamma(g) = (1-\gamma)\mathbb{E}[(\tau(X)-g(X))^2] - \gamma\,\mathbb{E}[\mathbf{1}(g(X)>0)\tau(X)]$. To make this objective differentiable, the paper replaces the indicator with a sigmoid $\sigma(\alpha(X)g(X))$, where $\alpha(X)>0$ is a learned covariate-dependent sharpness parameter. The function $\alpha$ is trained through a weighted binary-cross-entropy surrogate so that it becomes large where the current estimate already has the correct sign and small where the sign is wrong, thereby giving gradient signal exactly in decision-relevant regions. The second stage can use plug-in, RA, IPW, or DR pseudo-outcomes, and the DR version inherits a doubly robust error rate: the PT-CATE error is controlled as long as either the propensity score or the response functions are estimated well.

What would settle it

Recompute the quantity $M$ in the proof of Theorem 4.1 for a specific function class such as linear functions on $[0,1]$; since every positive constant belongs to the set $S_G$, the supremum $M$ is infinite and the sequence $\{\tau_n\}$ with norms approaching $M$ cannot be chosen, so the proof as written fails unless a boundedness assumption is added.

Watch

Extended reading notes

Core claim

The central claim is Theorem 4.1: for any fixed neural-network class $G$, there exists a true CATE $\tau^*$ outside $G$ such that every $L^2$-optimal approximation $g^*_{\tau^*}$ in $G$, when thresholded at zero, yields a policy with strictly lower value than the best thresholded policy available in $G$. Equivalently, minimizing mean-squared error of the CATE and maximizing the value of the induced thresholded policy are different objectives, and the former does not dominate the latter under model misspecification. The proposed remedy is the $\gamma$-PT-CATE loss $L_\gamma(g) = (1-\gamma)\mathbb{E}[(\tau(X)-g(X))^2] - \gamma\,\mathbb{E}[\mathbf{1}(g(X)>0)\tau(X)]$, whose minimizer interpolates between pure CATE estimation ($\gamma=0$) and pure policy-value maximization ($\gamma=1$). The paper shows consistency when pseudo-outcomes are used and an error bound with doubly robust nuisance rates, and it demonstrates empirically that increasing $\gamma$ reduces policy loss while only mildly worsening CATE error.

Load-bearing premise

The proof of the suboptimality theorem assumes a finite upper bound on the size of a certain family of step-function treatment effects, yet that family includes arbitrarily large constant functions, so the bound is not finite and the constructed counterexample does not go through.

Editorial extensions

If this is right

  • Two-stage CATE learners that target squared error can be suboptimal for thresholded decisions whenever the second-stage model class is misspecified, even if the estimators are otherwise state-of-the-art.
  • The $\gamma$-PT-CATE loss provides a continuous trade-off: $\gamma=0$ recovers standard CATE estimation, $\gamma=1$ becomes direct policy optimization, and intermediate values sacrifice a little CATE accuracy for better decisions.
  • The DR pseudo-outcome version of the PT-CATE loss carries a doubly robust guarantee, so the retargeted estimand can be learned reliably even if only one of the nuisance components is accurately estimated.
  • In the real-data experiment, the retargeted policy improves response probability by about 24.45% over a standard CATE-based policy at $\gamma=0.98$, with modest increases in estimated PEHE.
  • If the second-stage class is unrestricted and the true CATE lies in it, the method reduces to ordinary CATE estimation and introduces no additional bias.

Reading between the lines

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

  • Editorial inference: The same mismatch likely appears for other decision rules built from CATE estimates, such as top-k treatment allocation or cost-sensitive thresholds, wherever squared-error loss weights regions by outcome variance rather than by decision consequence.
  • Editorial inference: The learned sharpness parameter $\alpha(X)$ can be read as a soft confidence measure for the decision, suggesting a testable extension where $\alpha$ is reported alongside the thresholded policy to quantify decision uncertainty.
  • Editorial inference: A natural benchmark is to compare the PT-CATE curve of policy value versus PEHE against direct off-policy learning on datasets with strong misspecification; the paper's experiments suggest the trade-off curve dominates standard learners, but a systematic comparison across public benchmarks is not included.
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

5 major / 5 minor

Summary. The manuscript studies the common practice of thresholding CATE estimates to make binary treatment decisions. It argues that two-stage CATE learners, which minimize L2 error, may be suboptimal for decision-making when the second-stage model class is misspecified. It introduces the PT-CATE objective, a convex combination of L2 error and policy value, with an adaptive sigmoid approximation alpha(x), and proposes a three-step neural training algorithm. The paper claims three theoretical results: Theorem 4.1 (existence of CATEs for which L2-optimal CATE estimates are decision-suboptimal), Theorem 4.3 (consistency under oracle nuisances), and Theorem 4.4 (error rates under nuisance estimation error). Experiments on synthetic data and the Hillstrom dataset show improved policy loss for larger gamma at a modest PEHE cost.

Significance. The paper addresses a relevant and timely question: the disconnect between CATE accuracy and downstream decision performance. The PT-CATE objective is intuitive, the adaptive-sigmoid optimization trick is a reasonable practical proposal, and the empirical study is fairly thorough (four pseudo-outcomes, sample splitting, alternative nuisance models, and real-world data). The code is released, which is a concrete strength. If the theoretical claims are corrected, the paper would make a useful contribution. As it stands, however, the headline theoretical result is not established, so the theoretical contribution is currently limited.

major comments (5)
  1. [Appendix B.1, Eq. (16)] The proof of Theorem 4.1 is invalid as stated because M is infinite. S_G in Eq. (15) contains, as the proof itself notes, every positive constant step function. For a fixed-architecture network that can represent constants, the constant tau_c(x)=c is in S_G with ||tau_c||_infty=c for every c>0, so M := sup_{tau in S_G} ||tau||_infty = infinity. Consequently, the sequence in Eq. (17), the selection of n_epsilon in Eq. (18), and the construction in Eq. (24) of tau* with ||tau*||_infty > M are not well-defined. No boundedness assumption on tau or Y appears in Theorem 4.1, so this is a genuine gap in the central argument, not merely a missing regularity line.
  2. [Theorem 4.1 / Eq. (15)] Even if M were finite, the proof's move from tau* not in S_G to a set E of positive measure on which sign(g_{tau*}) differs from sign(tau*) does not follow. By Eq. (15), S_G requires sign agreement at every x; a function whose optimal approximation agrees except on a null set is already outside S_G. The theorem needs a definition of S_G in terms of sign agreement up to null sets, or an additional argument that the mismatch set has positive measure, because the strict policy-value inequality in Eq. (29) depends on it.
  3. [Theorem 4.1 statement] The theorem is stated for an arbitrary fixed-architecture network class G, but no non-degeneracy condition is given. For the degenerate class consisting only of constant functions, every L2-optimal estimate is constant, the induced policy is either treat-all or treat-none, and pi_{g*_tau} is optimal within Pi_G; the claimed strict inequality cannot hold. The statement needs to be restricted (for example, to classes G that contain a sign-changing function, or to a concrete class such as linear functions), or replaced by an existence result for a specific class that is sufficient for the paper's qualitative claim.
  4. [Theorem 4.4, Eq. (11)] The displayed bound contains the optimization-dependent term R^m. Since \hat g is defined as the minimizer of the estimated loss and g* is in G, R^m is non-positive if optimization is exact; if optimization is inexact, no control on R^m is supplied. The sentence in Section 4.3 that good nuisance estimation suffices for a good second-stage learner therefore does not follow from Eq. (11) alone. The theorem should state explicitly how R^m is handled, for example by assuming a bound on optimization error or by proving that R^m is negligible under the stated conditions.
  5. [Section 4.4 / Algorithm 1] Theorems 4.3 and 4.4 are stated for a fixed alpha, but Algorithm 1 alternately trains alpha (Step 2) and g (Step 3). The final g produced by the algorithm is a minimizer of a data-dependent loss in which alpha is no longer fixed. The paper does not provide a guarantee for this alternating procedure, so the theory and the algorithm are not directly connected. The authors should either extend the guarantees to the alternating scheme or explicitly state that the theory applies only to the fixed-alpha population objective and that the algorithm is a heuristic approximation.
minor comments (5)
  1. [Appendix B.1, Eq. (22)] The symbol tau*_{n_epsilon} is not defined and appears to be a typo for tau_{n_epsilon}.
  2. [Section 4.2, Eq. (8)] The binary cross-entropy term uses labels 1(tau(X)>0) and 1(tau(X)<0), but the case tau(X)=0 is not specified; this should be clarified, even if it is a measure-zero event under the assumed data-generating process.
  3. [Section 2] There is a typo in 'overregluarizing' in the discussion of Bonvini et al.; the text should read 'overregularizing'.
  4. [Section 5] The sentence 'We also compare the as compared to the behavioral policy' is ungrammatical and should be rewritten.
  5. [Theorem 4.3 proof, Eq. (35)] The step marked with 'proportional to' hides the constant E[(Y^m_eta - tau(X))^2] and the cancellation of the cross term via E[Y^m_eta - tau(X) | X]=0; this is correct, but the notation is sloppy and should be made explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's PT-CATE objective, consistency theorem, and error-rate analysis are derived from first principles; experiments use held-out ground-truth CATE or DR pseudo-outcomes; self-citations are not load-bearing.

full rationale

The paper does not fit its conclusions into its inputs. The central contribution, the PT-CATE objective in Eq. (5), is defined directly as a convex combination of squared CATE error and negative thresholded policy value, and the subsequent neural algorithm optimizes an adaptive approximation of that objective (Eqs. (6)-(9)). Theorem 4.3 is proved by the tower property using the fact that each pseudo-outcome has conditional mean tau(X), and Theorem 4.4/Theorem B.1 provides an explicit error-rate bound whose R^m term is an optimization residual, not a fitted constant disguised as a prediction. Experiments are evaluated on held-out ground-truth CATEs (simulated settings) or on DR pseudo-outcomes (Hillstrom), with gamma=0 as the standard CATE baseline, so the reported policy-loss improvements are not forced by construction. The paper cites several prior works by its own authors, but these citations appear in related-work discussion (e.g., [16,17,40,48]) and do not carry any central theorem, uniqueness claim, or ansatz; no load-bearing argument reduces to a self-citation. The possible objection that the proof of Theorem 4.1 uses an unbounded supremum M is a mathematical correctness concern about the proof as written, not a circularity: a flawed or missing proof is not the same as a derivation that is equivalent to its own inputs by definition. Under the proportionality and evidence rules, the honest finding is no significant circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The central theoretical claim depends on the standard causal assumptions, bounded nuisance conditions for the error-rate bound, a strong regularity condition in Theorem B.1, and an unstated boundedness assumption needed by the proof of Theorem 4.1. The method introduces gamma and the lower bound a as free parameters. No new physical entities are introduced beyond the PT-CATE estimand and the adaptive alpha function, both internal to the method.

free parameters (3)
  • gamma = scanned over {0, 0.2, 0.4, 0.8, 0.9, 0.98} in experiments
    The trade-off hyperparameter between CATE estimation error and policy value; the paper recommends choosing it by domain knowledge or validation loss, so it is a free parameter of the method.
  • a (lower bound for alpha) = tuned on validation set
    Section 4.2: 'The scalar a can be tuned by minimizing the loss from Eq. (5) on a validation set.' This is a tuned constant.
  • regularization parameter for second-stage g = custom per pseudo-outcome
    Appendix C: 'we choose a custom regularization parameter for each pseudo-outcome type that yields a misspecified initial CATE estimate.' This is a hand-chosen parameter in the experiments.
assumptions (4)
  • domain assumption Standard causal inference assumptions: consistency, overlap, ignorability (Assumption 3.1)
    Used to identify CATE and policy value from observational data; stated in Section 3.1.
  • domain assumption Bounded nuisance estimates and responses: p <= pi_hat <= 1-p and |mu_a| <= c
    Assumed in Theorem B.1 to bound the pseudo-outcome error terms; stated in Appendix B.3.
  • ad hoc to paper Regularity condition: there exists delta > 0 such that E[-Y^m sigma''(alpha gbar) alpha^2 (g_hat-g*)^2] / ||g*-g_hat||^2 >= delta
    Equation (39) in Appendix B.3, needed to control the second-order term in the Taylor expansion; it is a strong technical condition tailored to the proof.
  • ad hoc to paper Boundedness of the sign-correct CATE class so M is finite
    Required by the proof of Theorem 4.1 (Appendix B.1) but not stated in the theorem; without it the proof is invalid.
invented entities (2)
  • Policy-targeted CATE (PT-CATE)
    purpose: A new estimand defined as minimizer of the convex combination loss in Definition 4.2, balancing CATE MSE and policy value.
    PT-CATE is a definition introduced by the paper; no external falsifiable handle is provided.
  • Adaptive indicator approximation function alpha(X)
    purpose: A per-covariate steepness parameter in the sigmoid surrogate for the threshold indicator, learned via Eq. (8).
    It is part of the proposed algorithm and has no independent external evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Treatment Effect Estimation for Optimal Decision-Making." pith.science (2026). https://pith.science/paper/DQH4B2PU

@misc{pith2026250513092,
  author       = {Pith},
  title        = {Pith review of: Treatment Effect Estimation for Optimal Decision-Making},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DQH4B2PU}},
  note         = {Machine review of arXiv:2505.13092}
}
read the original abstract

Decision-making across various fields, such as medicine, heavily relies on conditional average treatment effects (CATEs). Practitioners commonly make decisions by checking whether the estimated CATE is positive, even though the decision-making performance of modern CATE estimators is poorly understood from a theoretical perspective. In this paper, we study optimal decision-making based on two-stage CATE estimators (e.g., DR-learner), which are considered state-of-the-art and widely used in practice. We prove that, while such estimators may be optimal for estimating CATE, they can be suboptimal when used for decision-making. Intuitively, this occurs because such estimators prioritize CATE accuracy in regions far away from the decision boundary, which is ultimately irrelevant to decision-making. As a remedy, we propose a novel two-stage learning objective that retargets the CATE to balance CATE estimation error and decision performance. We then propose a neural method that optimizes an adaptively-smoothed approximation of our learning objective. Finally, we confirm the effectiveness of our method both empirically and theoretically. In sum, our work is the first to show how two-stage CATE estimators can be adapted for optimal decision-making.

Figures

Figures reproduced from arXiv: 2505.13092 by the authors.

Figure 1
Figure 1. Illustrative example show￾ing the suboptimality of CATE estima￾tion for decision-making. The dotted lines show regularized two-stage CATE estimators. The blue line corresponds to standard two-stage CATE estimation, while the green and violet lines are gen￾erated by our method. The parameter γ quantifies the trade-off between CATE estimation error and decision-making performance. Details are in Sec. 5. Intuition: Why… view at source ↗
Figure 2
Figure 2. Experimental results for our proposed method with G being the class of linear models. Left: CATE estimator (blue) is the best linear approximation of the (nonlinear) ground-truth CATE (red). Center: the trained α(X) detects the region in which the estimated CATE has the wrong sign. Right: retargeted CATE estimators using our proposed loss with trained α(X) and different γ values. To provide an intuition on why two-s… view at source ↗
Figure 3
Figure 3. Overview of our second-stage architecture and our learning algorithm. • Step 1 (initial CATE estimation): train gθ by minimiz￾ing Lˆm γ=0,αϕ,ηˆ (g) over θ, using randomly initialized αϕ with ϕ frozen. This gives an initial CATE estimator. • Step 2 (region detection): train αϕ by minimizing Lˆm γ,gθ,ηˆ (αϕ) over ϕ, keeping θ frozen. The objective is for αϕ to iden￾tify covariate regions where gθ produces incorrect pr… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Experimental results for setting A. Shown: PEHE and policy loss over γ (lower = better). Shown: mean and standard errors over 5 runs. 0 0.2 0.4 0.8 0.9 0.98 0.100 0.075 0.050 0.025 0.000 0.025 0.050 0.075 0.100 PEHE Legend PEHE PI Policy loss PI Baseline = 0 0.990 0.98…
Figure 5
Figure 5. Figure 5: Experimental results for setting B. Shown: PEHE and policy loss over γ (lower = better). Shown: mean and standard errors over 5 runs. Experiments with estimated nuisance functions. We now consider two settings to analyze the effectiveness of our algorithm when using es…
Figure 6
Figure 6. Figure 6: Experimental results for real-world data. Shown: PEHE and policy loss over γ (lower = bet￾ter). Shown: Mean and 80% confi￾dence intervals over 5 runs. Results. We report PEHE and policy loss for all four pseudo￾outcomes (PI, RA, IPW, and DR) in [PITH_FULL_IMAGE:figure…
Figure 7
Figure 7. Figure 7: Results from [PITH_FULL_IMAGE:figures/full_fig_p023_7.png]
Figure 8
Figure 8. Figure 8: Results from [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 9
Figure 9. Figure 9: Experimental results for setting A with sample splitting. We re-ran our experiments from [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]
Figure 10
Figure 10. Figure 10: Experimental results for setting B with sample splitting. We re-ran our experiments from [PITH_FULL_IMAGE:figures/full_fig_p024_10.png]
Figure 11
Figure 11. Figure 11: Experimental results for setting A with TARNet. We re-run our experiments from [PITH_FULL_IMAGE:figures/full_fig_p025_11.png]
Figure 12
Figure 12. Figure 12: Experimental results for setting B with TARNet. We re-run our experiments from [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 58 canonical work pages

  1. [1]

    Lifetime earnings and the vietnam era draft lotter: Evidence from social security administrative records

    Joshua D. Angrist. “Lifetime earnings and the vietnam era draft lotter: Evidence from social security administrative records”. In:The American Economic Review 80.3 (1990), pp. 313–336

    work page 1990

  2. [2]

    Policy learning with observational data

    Susan Athey and Stefan Wager. “Policy learning with observational data”. In:Econometrica 89.1 (2021), pp. 133–161

    work page 2021

  3. [3]

    Efficient policy learning from surrogate-loss classification reductions

    Andrew Bennett and Nathan Kallus. “Efficient policy learning from surrogate-loss classification reductions”. In: ICML. 2020

    work page 2020

  4. [4]

    More efficient off-policy evaluation through regularized targeted learning

    Aurelien Bibaut et al. “More efficient off-policy evaluation through regularized targeted learning”. In: ICML. 2019

    work page 2019

  5. [5]

    Estimating counterfactual treatment outcomes over time through adversarially balanced representations

    Ioana Bica et al. “Estimating counterfactual treatment outcomes over time through adversarially balanced representations”. In: ICLR. 2020

    work page 2020

  6. [6]

    Minimax optimal subgroup identifi- cation

    Matteo Bonvini, Edward H. Kennedy, and Luke J. Keele. “Minimax optimal subgroup identifi- cation”. In: arXiv preprint arXiv:2306.17464 (2023)

    arXiv 2023

  7. [7]

    Double/debiased machine learning for treatment and structural parameters

    Victor Chernozhukov et al. “Double/debiased machine learning for treatment and structural parameters”. In: The Econometrics Journal 21.1 (2018), pp. C1–C68

    work page 2018

  8. [8]

    Nonparametric estimation of heterogeneous treat- ment effects: From theory to learning algorithms

    Alicia Curth and Mihaela van der Schaar. “Nonparametric estimation of heterogeneous treat- ment effects: From theory to learning algorithms”. In: AISTATS. 2021

    work page 2021

Show all 61 references
  1. [9]

    On inductive biases for heterogeneous treatment effect estimation

    Alicia Curth and Mihaela van der Schaar. “On inductive biases for heterogeneous treatment effect estimation”. In: NeurIPS. 2021

    2021

  2. [10]

    Doubly robust policy evaluation and learning

    Miroslav Dudik, John Langford, and Lihong Li. “Doubly robust policy evaluation and learning”. In: ICML. 2011

    2011

  3. [11]

    Inferring effect ordering without causal effect estimation

    Carlos Fernández-Loría and Jorge Loría. “Inferring effect ordering without causal effect estimation”. In: arXiv preprint arXiv:2206.12532 (2024)

    2024

  4. [12]

    Causal classification: Treatment effect estimation vs. outcome prediction

    Carlos Fernández-Loría and Foster Provost. “Causal classification: Treatment effect estimation vs. outcome prediction”. In: Journal of Machine Learning Research 23.59 (2022), pp. 1–35

    2022

  5. [13]

    Causal decision making and causal effect esti- mation are not the same... and why it matters

    Carlos Fernández-Loría and Foster Provost. “Causal decision making and causal effect esti- mation are not the same... and why it matters”. In: INFORMS Journal on Data Science 1.1 (2022), pp. 4–16

    2022

  6. [14]

    Causal machine learning for predicting treatment outcomes

    Stefan Feuerriegel et al. “Causal machine learning for predicting treatment outcomes”. In: Nature Medicine (2024)

    2024

  7. [15]

    Orthogonal statistical learning

    Dylan J. Foster and Vasilis Syrgkanis. “Orthogonal statistical learning”. In: The Annals of Statistics 53.3 (2023), pp. 879–908

    2023

  8. [16]

    Model-agnostic meta-learners for estimating heterogeneous treatment effects over time

    Dennis Frauen, Konstantin Hess, and Stefan Feuerriegel. “Model-agnostic meta-learners for estimating heterogeneous treatment effects over time”. In: ICLR. 2025

    2025

  9. [17]

    Fair off-policy learning from observational data

    Dennis Frauen, Valentyn Melnychuk, and Stefan Feuerriegel. “Fair off-policy learning from observational data”. In: ICML. 2024

    2024

  10. [18]

    Causal inference in public health

    Thomas A. Glass et al. “Causal inference in public health”. In:Annual Review of Public Health 34 (2013), pp. 61–75

    2013

  11. [19]

    Learning disentangled representations for counterfac- tual regression

    Negar Hassanpour and Russell Greiner. “Learning disentangled representations for counterfac- tual regression”. In: ICLR. 2020

    2020

  12. [20]

    Bayesian nonparametric modeling for causal inference

    Jennifer L. Hill. “Bayesian nonparametric modeling for causal inference”. In: Journal of Computational and Graphical Statistics 20.1 (2011), pp. 2017–2040

    2011

  13. [21]

    Doubly robust off-policy value evaluation for reinforcement learning

    Nan Jiang and Lihong Li. “Doubly robust off-policy value evaluation for reinforcement learning”. In: ICML. 2016

    2016

  14. [22]

    Learning representations for counter- factual inference

    Fredrik D. Johansson, Uri Shalit, and David Sonntag. “Learning representations for counter- factual inference”. In: ICML. 2016

    2016

  15. [23]

    Balanced policy evaluation and learning

    Nathan Kallus. “Balanced policy evaluation and learning”. In: NeurIPS. 2018

    2018

  16. [24]

    More efficient policy learning via optimal retargeting

    Nathan Kallus. “More efficient policy learning via optimal retargeting”. In: Journal of the American Statistical Association 116.534 (2021), pp. 646–658

    2021

  17. [25]

    Double reinforcement learning for efficient off-policy evaluation in markov decision processes

    Nathan Kallus and Masatoshi Uehara. “Double reinforcement learning for efficient off-policy evaluation in markov decision processes”. In: Journal of Machine Learning Research 21 (2020), pp. 1–63

    2020

  18. [26]

    Doubly robust off policy value and gradient estimation for deterministic policies

    Nathan Kallus and Masatoshi Uehara. “Doubly robust off policy value and gradient estimation for deterministic policies”. In: NeurIPS. 2020. 11

    2020

  19. [27]

    Efficiently breaking the curse of horizon in off- policy evaluation with double reinforcement learning

    Nathan Kallus and Masatoshi Uehara. “Efficiently breaking the curse of horizon in off- policy evaluation with double reinforcement learning”. In: Operations Research 70.6 (2022), pp. 3282–3302

    2022

  20. [28]

    Intrinsically efficient, stable, and bounded off-policy evaluation for reinforcement learning

    Nathan Kallus and Masatoshi Uehara. “Intrinsically efficient, stable, and bounded off-policy evaluation for reinforcement learning”. In: NeurIPS. 2019

    2019

  21. [29]

    Policy evaluation and optimization with continuous treat- ments

    Nathan Kallus and Angela Zhou. “Policy evaluation and optimization with continuous treat- ments”. In: AISTATS. 2018

    2018

  22. [30]

    Doubly robust distributionally roust off-policy evaluation and learning

    Nathan Kallus et al. “Doubly robust distributionally roust off-policy evaluation and learning”. In: ICML. 2022

    2022

  23. [31]

    Learning to rank for optimal treatment allocation under resource constraints

    Fahad Kamran, Maggie Maker, and Jenna Wiens. “Learning to rank for optimal treatment allocation under resource constraints”. In: AISTATS. 2024

    2024

  24. [32]

    Towards optimal doubly robust estimation of heterogeneous causal effects

    Edward H. Kennedy. “Towards optimal doubly robust estimation of heterogeneous causal effects”. In: Electronic Journal of Statistics 17.2 (2023), pp. 3008–3049

    2023

  25. [33]

    Semiparametric coun- terfactual density estimation

    Edward H. Kennedy, Sivaraman Balakrishnan, and Larry Wasserman. “Semiparametric coun- terfactual density estimation”. In: Biometrika (2023)

    2023

  26. [34]

    Fair and robust estimation of heterogeneous treatment effects for policy learning

    Kwangho Kim and José R. Zubizarreta. “Fair and robust estimation of heterogeneous treatment effects for policy learning”. In: ICML. 2023

    2023

  27. [35]

    Adam: A method for stochastic optimization

    Diederik P. Kingma and Jimmy Ba. “Adam: A method for stochastic optimization”. In: ICLR. 2015

    2015

  28. [36]

    Data-driven allocation of preventive care with application to diabetes mellitus type II

    Mathias Kraus, Stefan Feuerriegel, and Maytal Saar-Tsechansky. “Data-driven allocation of preventive care with application to diabetes mellitus type II”. In: Manufacturing & Service Operations Management) 26.1 (2024), pp. 137–153

    2024

  29. [37]

    Metalearners for estimating heterogeneous treatment effects using machine learning

    Sören R. Künzel et al. “Metalearners for estimating heterogeneous treatment effects using machine learning”. In: Proceedings of the National Academy of Sciences (PNAS) 116.10 (2019), pp. 4156–4165

    2019

  30. [38]

    Causal machine learning for cost-effective allocation of development Aid

    Milan Kuzmanovic et al. “Causal machine learning for cost-effective allocation of development Aid”. In: KDD. 2024

    2024

  31. [39]

    Forecasting treatment responses over time using recurrent marginal structural networks

    Bryan Lim, Ahmed M. Alaa, and Mihaela van der Schaar. “Forecasting treatment responses over time using recurrent marginal structural networks”. In: NeurIPS. 2018

    2018

  32. [40]

    Causal transformer for estimat- ing counterfactual outcomes

    Valentyn Melnychuk, Dennis Frauen, and Stefan Feuerriegel. “Causal transformer for estimat- ing counterfactual outcomes”. In: ICML. 2022

    2022

  33. [41]

    On a general class of orthogonal learners for the estimation of heterogeneous treatment effects

    Pawel Morzywolek, Johan Decruyenaere, and Stijn Vansteelandt. “On a general class of orthogonal learners for the estimation of heterogeneous treatment effects”. In: arXiv preprint arXiv:2303.12687 (2023)

    2023 arXiv

  34. [42]

    Marginal mean models for dynamic regimes

    Susan Murphy, Mark van der Laan, and James M. Robins. “Marginal mean models for dynamic regimes”. In: Journal of the American Statistical Association 96.456 (2001), pp. 1410–1423

    2001

  35. [43]

    Optimal dynamic treatment regimes

    Susan A. Murphy. “Optimal dynamic treatment regimes”. In: Journal of the Royal Statistical Society: Series B 65.2 (2003), pp. 331–355

    2003

  36. [44]

    Fair inference on outcomes

    Razieh Nabi and Ilya Shpitser. “Fair inference on outcomes”. In: AAAI. 2018

    2018

  37. [45]

    Quasi-oracle estimation of heterogeneous treatment effects

    Xinkun Nie and Stefan Wager. “Quasi-oracle estimation of heterogeneous treatment effects”. In: Biometrika 108.2 (2021), pp. 299–319

    2021

  38. [46]

    Performance guarantees for individualized treatment rules

    Min Qian and Susan A. Murphy. “Performance guarantees for individualized treatment rules”. In: Annals of Statistics 39.2 (2011), pp. 1180–1210

    2011

  39. [47]

    Estimating causal effects of treatments in randomized and nonrandomized studies

    Donald B. Rubin. “Estimating causal effects of treatments in randomized and nonrandomized studies”. In: Journal of Educational Psychology 66.5 (1974), pp. 688–701

    1974

  40. [48]

    Reliable off-policy learning for dosage combinations

    Jonas Schweisthal et al. “Reliable off-policy learning for dosage combinations”. In: NeurIPS. 2023

    2023

  41. [49]

    Estimating individual treatment effect: Generalization bounds and algorithms

    Uri Shalit, Fredrik D. Johansson, and David Sontag. “Estimating individual treatment effect: Generalization bounds and algorithms”. In: ICML. 2017

    2017

  42. [50]

    Counterfactual risk minimization: Learning from logged bandit feedback

    Adith Swaminathan and Thorsten Joachims. “Counterfactual risk minimization: Learning from logged bandit feedback”. In: ICML. 2015

    2015

  43. [51]

    Data-efficient off-Policy policy evaluation for rein- forcement Learning

    Phillip S. Thomas and Emma Brunskill. “Data-efficient off-Policy policy evaluation for rein- forcement Learning”. In: ICML. 2016

    2016

  44. [52]

    Interpretable off-policy learning via hyperbox search

    Daniel Tschernutter, Tobias Hatt, and Stefan Feuerriegel. “Interpretable off-policy learning via hyperbox search”. In: ICML. 2022. 12

    2022

  45. [53]

    Statistical inference for variable importance

    Mark J. van der Laan. “Statistical inference for variable importance”. In: The International Journal of Biostatistics 2.1 (2006), pp. 1–31

    2006

  46. [54]

    Targeted maximum likelihood learning

    Mark J. van der Laan and Donald B. Rubin. “Targeted maximum likelihood learning”. In: The International Journal of Biostatistics 2.1 (2006)

    2006

  47. [55]

    Asymptotic statistics

    Aart van der Vaart. Asymptotic statistics. Cambridge: Cambridge University Press, 1998

    1998

  48. [56]

    Causal inference in economics and marketing

    Hal R. Varian. “Causal inference in economics and marketing”. In:Proceedings of the National Academy of Sciences (PNAS) 113.27 (2016), pp. 7310–7315

    2016

  49. [57]

    Estimation and inference of heterogeneous treatment effects using random forests

    Stefan Wager and Susan Athey. “Estimation and inference of heterogeneous treatment effects using random forests”. In: Journal of the American Statistical Association 113.523 (2018), pp. 1228–1242

    2018

  50. [58]

    C-learning: A new classification framework to estimate optimal dynamic treatment regimes

    Baqun Zhang and Min Zhang. “C-learning: A new classification framework to estimate optimal dynamic treatment regimes”. In: Biometrics 74.3 (2018), pp. 891–899

    2018

  51. [59]

    Estimating optimal treatment regimes from a classification perspective

    Baqun Zhang et al. “Estimating optimal treatment regimes from a classification perspective”. In: Stat 1.1 (2012), pp. 103–112

    2012

  52. [60]

    Learning overlapping representations for the estimation of individualized treatment effects

    Yao Zhang, Alexis Bellot, and Mihaela van der Schaar. “Learning overlapping representations for the estimation of individualized treatment effects”. In: AISTATS. 2020

    2020

  53. [61]

    Counterfactual prediction for outcome-oriented treatments

    Hao Zou et al. “Counterfactual prediction for outcome-oriented treatments”. In: ICML. 2022. 13 A Extended related work A.1 Deep learning for CATE estimation In recent years, deep neural networks have gained considerable traction for estimating the conditional average treatment...

    2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.