Local conformal prediction for individual causal effects

Fernando Delbianco; Fernando Tohm\'e

REVIEW 4 major objections 5 minor 18 references

Local conformal prediction for individual causal effects

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Local conformal intervals can cover individual causal effects with finite-sample guarantees.

desk verdict The paper's core coverage theorem rests on a selection measurability assumption that the implementation violates twice over, so the headline guarantee is unsupported. read the letter →

arxiv 2608.09612 v1 pith:FEYI4YN2 submitted 2026-08-10 stat.ME econ.EM

Fernando Delbianco , Fernando Tohmé This is my paper · ORCID

classification stat.MEecon.EM MSC 62G1562D2062G08

keywords conformalpredictionindividualtreatmenteffectscausalinferenceForestsheterogeneousdoublyrobustestimationlocalcalibrationfinite-samplecoverage

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 claims that a prediction interval for the causal effect on one specific individual—the difference between that person's treated and untreated outcomes—can be made finite-sample valid, not just asymptotically valid for a conditional average. The method builds a calibration set from units close to the query in the causally relevant covariate subspace, adds synthetic local units, filters by a score that depends only on covariates and treatment, and applies split-conformal calibration with doubly robust AIPW scores. If the selection condition holds, Theorem 1 gives $P(\tau_q \in \hat{C}(\tau_q)) \ge 1-\alpha$, and the local design moves coverage from merely marginal toward conditional at the query point. The empirical sections show local calibration beating a global baseline on point accuracy while keeping coverage at or above nominal, which matters for personalized decisions under strong treatment-effect heterogeneity.

What carries the argument

The object carrying the argument is the super-relevant calibration set $R_q^{*(t)}$, obtained by taking the top fraction of treated (or control) units by Causal-Forest-weighted cosine similarity to $x_q$, adding Gaussian-perturbed synthetic copies with local linear imputation, and retaining the upper half by the composite score $\phi_i(q) = (1-\lambda)\phi_i^{\mathrm{Sh}}(q) + \lambda \phi_i^{\mathrm{PS}}(q)$, where $\phi_i^{\mathrm{Sh}}$ is a Shapley influence weight from Causal Forest leaf structures and $\phi_i^{\mathrm{PS}} = 1-|\hat{e}(X_i)-\hat{e}(x_q)|$ is propensity-score proximity. The design intends $\phi_i(q)$ to be a function of $(X_i,T_i,X_q)$ only, so membership in the calibration set is outcome-independent; split-conformal exchangeability of AIPW conformity scores $V_i = |\hat{\Gamma}_i - \hat{\tau}_{\mathrm{OOB}}(X_i)|$ then yields the finite-sample quantile $Q_{1-\alpha}$ that defines the interval. The AIPW pseudo-outcome $\hat{\Gamma}_i$ is doubly robust and Neyman-orthogonal, so first-order nuisance errors do not shift the scores to leading order.

What would settle it

Permute the observed outcomes $Y$ among units while keeping $(X,T)$ fixed, rerun the full pipeline's Stage 1 (variable importance) and Stage 2 (synthetic imputation), and compute leave-one-out coverage at the nominal level: if calibration selection truly ignores outcomes, coverage stays at $1-\alpha$, whereas any significant drop demonstrates outcome dependence in the calibration set and directly violates Assumption 3.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that individual causal effects admit locally calibrated conformal intervals: under SUTVA, strong ignorability, and a selection measurability condition (the composite score $\phi_i(q)$ used to enter $R_q^{*(t)}$ depends only on $(X_i,T_i,X_q)$, not on any outcome), the interval $\hat{C}(\tau_q)$ from Section 3.2 satisfies $P(\tau_q \in \hat{C}(\tau_q)) \ge 1-\alpha$ in finite samples. The same conformal mechanism that covers the counterfactual $Y_q(1)$ shifts by the observed $Y_q(0)$ to cover the individual effect $\tau_q$. Because calibration is restricted to a causally relevant neighborhood instead of the full sample, the paper argues the marginal guarantee becomes approximately conditional at $x_q$, a strictly stronger statement than global reweighting approaches provide.

Load-bearing premise

The coverage guarantee collapses if the rule for choosing the calibration set uses outcome values; the paper's Assumption 3 forbids this, but the implementation computes Causal Forest variable importance and imputed synthetic outcomes from $Y$, so the guarantee holds only insofar as those outcome-dependent quantities do not actually break exchangeability of the conformity scores.

Editorial extensions

If this is right

A practitioner can report an interval for the effect of a treatment on the particular individual in front of them, with a finite-sample $1-\alpha$ coverage guarantee, rather than an interval for the population conditional mean.
Local calibration gives approximately conditional coverage at the query point, so the reported uncertainty is more relevant to that unit than a global marginal interval.
The doubly robust AIPW scores let the method keep coverage while centering the interval on a semiparametrically efficient point estimate.
On the synthetic and IHDP benchmarks, local strategies achieve lower RMSE and MAE than the global baseline while maintaining nominal or above-nominal coverage, including 100% coverage on IHDP.
The framework is modular: any calibration-set selector that is outcome-independent and any exchangeable conformity scores can replace the specific Causal Forest and AIPW implementation.

Reading between the lines

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

Extension (not the paper's claim): the current implementation's Stage 1 variable importance and Stage 2 synthetic outcomes are fitted on $Y$, so a direct test is to rerun the pipeline with outcome-permuted data and compare leave-one-out coverage; if selection truly ignored outcomes, coverage would be unchanged.
Extension (not the paper's claim): the selection-measurability principle suggests that any outcome-blind comparator rule—propensity matching, representation learning without labels, or instrument-based similarity—could be conformalized for individual effects, including instrumental-variable settings the paper mentions only as future work.
Extension (not the paper's claim): the approximate-conditional-coverage claim is testable by stratifying queries into covariate regions and checking coverage within each region; the gap between region-level and nominal coverage should shrink as the local mechanism becomes more homogeneous.
Extension (not the paper's claim): synthetic augmentation's benefit is tied to local positivity, so a diagnostic is to vary the augmentation noise scale and the number of synthetic points and observe whether coverage and width respond; if the local sample is already large, augmentation should be neutral.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Dear colleague,

This paper is worth a look for the problem it tackles, but the main guarantee doesn't survive contact with the implementation. The idea is to build conformal prediction intervals for individual treatment effects by calibrating on a local, causally relevant neighborhood rather than the whole dataset. That's a legitimate gap, and the local weighting, synthetic augmentation, and AIPW scores form a plausible recipe. The writing is clear and the experiments, such as they are, show modest gains in point prediction.

The soft spot is load-bearing. Theorem 1 requires Assumption 3, that the calibration set be selected without using outcomes. The text says the assumption is satisfied by construction, but it is not. Stage 1 computes variable importance omega_j from a Causal Forest fitted on Y, and that importance re-weights the cosine similarity that defines the relevant neighborhood. Stage 2 imputes synthetic outcomes from a local model fitted on Y, so the synthetic calibration points carry outcome information. Either way, membership in R* depends on Y, and exchangeability of the conformity scores fails.

There's a second, independent problem. The predictive model mu_hat is fitted on the same calibration set R* used to form the non-conformity scores. Those in-sample residuals are systematically smaller than the query's out-of-sample residual, so the conformal quantile is biased low. The proof treats mu_hat as fixed, but it is estimated on the calibration data. This is not a minor technicality; it is the difference between a valid finite-sample guarantee and an empirical heuristic.

The empirical coverage numbers, which are all at or above nominal, are consistent with these artifacts rather than evidence for the theorem. The paper itself acknowledges the modular requirements at the end of Section 7: selection without outcomes and exchangeable scores. It just doesn't meet them.

What's genuinely useful is the local-calibration framing. I could see a fixed version working: choose the neighborhood with pre-outcome variable importance (e.g., unsupervised or split-sample), set aside a clean calibration fold, and use proper split conformal with separate training. The current manuscript doesn't do that.

I'd send this to a referee for a major revision, not because the current proof works but because the problem is real and the fix is identifiable. If the authors can re-do the selection and calibration to satisfy their own assumptions, there's a publishable paper here. As is, it's a heuristic with an unproven guarantee.

Referee Report

4 major / 5 minor

Summary. The paper proposes an Individualized Causal Prediction (ICP) framework that uses conformal prediction to construct finite-sample prediction intervals for an individual causal effect tau_q = Y_q(1) - Y_q(0). The method localizes the calibration set to a neighborhood of the query unit using cosine similarity weighted by Causal Forest variable importance, augments that neighborhood with synthetic observations, filters via a composite "super-relevance" score, and calibrates with AIPW pseudo-outcome conformity scores. The central theoretical claim is Theorem 1, which asserts marginal coverage at level 1 - alpha under SUTVA, strong ignorability, and an outcome-independent calibration-set selection condition (Assumption 3). The paper reports experiments on a synthetic benchmark and the IHDP dataset, claiming that local strategies improve point accuracy over global baselines while maintaining nominal or above-nominal coverage.

Significance. If Theorem 1 were valid, the paper would address a genuinely important gap: standard CATE intervals cover conditional means but not individual counterfactual realizations, and local conformal calibration is a natural route to more conditional coverage. The paper also deserves credit for explicitly stating a selection measurability condition and for recognizing that outcome-dependent calibration-set construction would break exchangeability. The modular framing and the use of AIPW scores are reasonable building blocks. However, the manuscript's central guarantee is not established: the implementation violates the very assumption on which the proof relies, and the conformal scores are computed in-sample rather than in a split-conformal manner. Because the main theorem is the paper's core contribution, these are load-bearing flaws rather than presentation issues.

major comments (4)

[Section 4.1, Stages 1-3; Assumption 3] Assumption 3 is violated by the implementation. The cosine-similarity weights omega_j come from a global Causal Forest fitted on observed Y; the Shapley influence weights phi_i^Sh are derived from Causal Forest leaf weights of a forest fitted on Y; and the synthetic outcomes Ytilde_k are imputed from a local linear model fitted on Y. Since R* is formed by filtering the union of R_q and S_q on phi_i(q), membership in R* depends on outcome values. The proof's statement that 'calibration-set membership is independent of the outcome values' therefore does not hold for the algorithm as implemented, and the exchangeability step in Theorem 1 collapses.
[Section 3.2; proof of Theorem 1] The non-conformity scores V_i = |Y_i - muhat(X_i)| are computed on the same set R* that is used to fit muhat. The proof declares that muhat is treated as fixed, but muhat is fitted on R* before the scores are formed. In-sample residuals are not exchangeable with the out-of-sample query score V_q; they are systematically smaller, so the conformal quantile is biased downward. Split-conformal validity requires the score function to be fitted on a separate fold or to be out-of-fold, and the paper does neither.
[Section 3.3 vs. Section 4, Stage 4] Theorem 1 is proved for residual scores |Y_i - muhat(X_i)|, but the implementation uses AIPW pseudo-outcome scores |Gamma_i - tauOOB(X_i)|. The paper provides no theorem showing that these AIPW scores, after local filtering and soft weighting, are exchangeable with the corresponding query score. The mismatch means the finite-sample coverage claim does not cover the actual procedure evaluated in the experiments.
[Section 4.1, Stage 2; Section 3.2, stage B] The synthetic augmentation step generates Xtilde by Gaussian perturbation and imputes Ytilde from a local linear model fitted on observed Y. These synthetic points are not draws from the true conditional distribution of Y given X and T, so they cannot be treated as exchangeable calibration observations. The paper appeals to 'local ergodicity of the SCM' but does not formalize this assumption or prove that imputed points preserve the rank condition underlying conformal coverage. Synthetic augmentation is therefore not a valid device for restoring small local sample sizes within the stated theorem.

minor comments (5)

[Abstract; Section 1] The abstract contains a typo: 'SUTV A' should be 'SUTVA'.
[Section 3] The headings '3.2 Conformalization of the Individual Causal Effect' and '3.3 Coverage Guarantee' each appear twice, which disrupts the numbering and makes the logical structure harder to follow.
[Section 7] The Discussion states that any valid implementation must select the calibration set without using outcomes, but Stage 2's synthetic outcome imputation uses Y directly. The text should explicitly acknowledge this conflict with the formal condition rather than presenting it as automatically satisfied.
[Tables 1-2] The sentence in Section 5.2 that coverage 'confirms the finite-sample validity guarantee' overstates what an empirical coverage proportion can show, particularly without standard errors or repeated-seed variation; at best these tables illustrate that the intervals are conservative in the studied settings.
[References] The Cunningham (2026) citation is to a Substack blog post; it is not a peer-reviewed source and should be replaced or removed if a citable reference is needed for the epigraph.

Circularity Check

3 steps flagged · score 8.0 of 10

Theorem 1's coverage guarantee assumes away the outcome-dependence that the algorithm's own implementation introduces, and synthetic calibration data are generated from the same outcome-fitted model, so the claimed finite-sample validity does not follow.

fitted input called prediction [Section 3.3, Assumption 3 and Theorem 1 proof; Section 4.1, Stages 1 and 3]
"Assumption 3 (Selection Measurability). The composite super-relevance score φ_i(q) used to form R∗(t)_q is a function of (X_i, T_i, X_q) only, and does not depend on any outcome value Y_j ... Assumption 3 is satisfied by construction: φ^Sh_i(q) is derived from Causal Forest leaf weights (functions of X only) ... Stage 1: ... variable importance scores ω_j from a global Causal Forest."

Theorem 1's proof relies on Assumption 3: 'By Assumption 3, the event {i ∈ R∗(1)_q} depends only on (X, T), so calibration-set membership is independent of the outcome values.' But the implemented selection rule is outcome-dependent: the cosine-similarity weights ω_j come from a global Causal Forest, and the Shapley influence weights φ^Sh_i(q) also come from a Causal Forest; a Causal Forest is fitted on observed outcomes Y, as the AIPW formula in Stage 4 makes explicit through Y_i. Hence R∗ membership depends on Y, contradicting Assumption 3. The coverage guarantee is derived from an assumption that the algorithm's own construction violates; the 'outcome-independent' property is asserted by definition rather than achieved, making the theorem circular relative to the implemented method.
fitted input called prediction [Section 3.2, stage B; Section 4.1, Stage 2]
"S^(t)_q = {~Z_k = (~X_k, T_k=t, ~Y_k)}^m_{k=1}, ~Z_k ∼ P(Z|Z∈R^(t)_q). The generative model P is detailed in Section 4. ... Outcome values ~Y_k are imputed from a local linear model."

Synthetic observations are added to the calibration set and their outcomes are generated by a local linear model fitted on observed Y. These synthetic units then enter R∗ and contribute non-conformity scores V_i = |Y_i − µˆ(t)(X_i)|. The calibration score distribution therefore consists largely of in-sample residuals of the imputation model, not of genuine counterfactual outcomes. The conformal quantile is a quantile of model-generated residuals, so using it to cover a real query's potential outcome is circular: the calibration evidence is manufactured from the same fitted outcome model it is supposed to validate. Exchangeability between synthetic calibration scores and the true query score is not established, since the query's outcome is not drawn from the imputation model.

1 more flagged steps

self definitional [Section 3.2 (i)-(ii); Section 3.3, Theorem 1 proof; Section 4.1, Stages 3-4]
"(i) A predictive model µˆ(t) is fitted on R∗(t)_q. (ii) For each i ∈ R∗(t)_q, the non-conformity score is V^(t)_i = |Y_i − µˆ(t)(X_i)|. ... Treating the estimated nuisance function µˆ(1) as fixed, the hypothetical score for the query is V_q = |Y_q(1) − µˆ(1)(x_q)|."

Split-conformal validity requires the calibration scores to be computed with a model that is fixed before seeing the calibration outcomes. Here the same set R∗ is used both to fit the model and to produce the calibration residuals: Stage 4 states 'the training set consists of the top 50% by φ_i(q), with scores used as soft sample weights in the Causal Forest fit,' and then computes V_i for 'each observation i in the training set.' The proof's 'treating µˆ as fixed' is only legitimate if µˆ was trained on an independent fold; in this implementation the model and the calibration scores come from the same observations. The exchangeability of (V_1,...,V_n,V_q) is therefore assumed rather than established, so the finite-sample coverage theorem does not apply to the implemented procedure.

full rationale

The paper's central claim is a finite-sample conformal coverage guarantee for individual causal effects. The proof of Theorem 1 is internally valid only under Assumption 3, which requires that the calibration set be selected without any outcome information, and under the standard split-conformal requirement that the scoring model be fixed before calibration. Both conditions are violated by the paper's own implementation. Stage 1 uses Causal Forest variable importance estimated on Y to define the cosine-similarity neighborhood; Stage 3 uses Shapley influence weights from a Causal Forest fit on Y; and Stage 2 imputes synthetic outcomes from a local linear model fitted on Y, making the calibration set outcome-dependent in multiple ways. Additionally, the same R∗ serves as both the training set for the Causal Forest and the source of the conformity scores, so the proof's 'treating µˆ as fixed' does not hold. The empirical coverages in Tables 1-2 are therefore not evidence for the stated finite-sample guarantee; they reflect artifacts of in-sample or model-generated scores. This is a substantial circularity in the derivation chain, not merely a minor self-citation issue. A hypothetical procedure that selected R∗ purely from X and T and used an independently pre-fitted model could satisfy the theorem, but that is not the algorithm described and evaluated.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The coverage theorem rests on causal identification assumptions plus Assumption 3, which the implementation violates because outcome-fitted variable importance and imputed synthetic outcomes affect calibration selection. The method also relies on hand-chosen hyperparameters. No new physical or causal entities are introduced.

free parameters (7)

kappa (top percentage for relevant set R_q) = 60
Chosen by hand; controls neighborhood size and the exchangeability versus relevance tradeoff. Not derived from data or theory.
m (synthetic sample size) = 500
Chosen by hand; affects calibration size and coverage behavior when synthetic points are included.
sigma_synth (Gaussian perturbation scale) = 0.1
Chosen by hand; controls the spread of synthetic neighbors around observed ones.
lambda (Shapley versus propensity weight) = 0.4
Chosen by hand; blends the two components of the super-relevance score.
minimum observations per arm = 40
Chosen by hand; prevents the relevant set from being too small.
propensity keep interval for synthetic points = [0.05, 0.95]
Chosen by hand; discards synthetic units with extreme predicted propensity.
top quantile for super-relevant set = 0.50
Chosen by hand; final calibration set retains the upper half by composite score.

assumptions (6)

domain assumption SUTVA (Assumption 1)
Potential outcomes are well-defined and no interference; standard causal identification assumption.
domain assumption Strong ignorability and overlap (Assumption 2)
Unconfoundedness plus bounded propensity scores; needed to relate observed outcomes to potential outcomes.
ad hoc to paper Selection measurability (Assumption 3)
The paper asserts the calibration set depends only on (X,T), but the implementation uses outcome-fitted variable importance and imputed synthetic outcomes, so this assumption is violated by construction.
standard math Exchangeability of conformity scores under selection
Split-conformal validity requires the calibration scores and query score to be exchangeable; the proof invokes this without establishing it for synthetic points.
ad hoc to paper Local ergodicity of the SCM for synthetic augmentation
Section 3.2 B assumes synthetic draws from the estimated local conditional distribution behave like real calibration units; this is not among the formal assumptions for Theorem 1.
ad hoc to paper Nuisance functions treated as fixed in the proof
Theorem 1's proof says treating the estimated nuisance function as fixed, but finite-sample coverage requires accounting for estimation of the model used to define scores.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Local conformal prediction for individual causal effects." pith.science (2026). https://pith.science/paper/FEYI4YN2

@misc{pith2026260809612,
  author       = {Pith},
  title        = {Pith review of: Local conformal prediction for individual causal effects},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FEYI4YN2}},
  note         = {Machine review of arXiv:2608.09612}
}

read the original abstract

Standard CATE estimators become inadequate under strong treatment-effect heterogeneity: confidence intervals for conditional means need not cover individual counterfactual effects. We propose an Individualized Causal Prediction (ICP) framework that constructs finite-sample valid conformal prediction intervals for the individual causal effect of a specific query unit. The method localizes calibration to a causally relevant neighborhood using cosine similarity weighted by Causal Forest variable importance, augments small local samples synthetically, and calibrates intervals with doubly robust AIPW conformity scores satisfying Neyman orthogonality. Under standard identifying assumptions (SUTVA and strong ignorability) and an outcome-independent calibration-set selection condition, the resulting intervals attain marginal coverage at the nominal level. The local design also supports approximately conditional coverage by making calibration scores more representative of the query unit. Experiments on a high-heterogeneity synthetic dataset and the IHDP benchmark demonstrate that local strategies improve point accuracy over global baselines while maintaining nominal or above-nominal coverage.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

18 extracted references · 8 canonical work pages

[1]

2026 , month = may, howpublished =

Cunningham, Scott , title =. 2026 , month = may, howpublished =

work page 2026
[2]

Journal of the American Statistical Association , year =

Wager, Stefan and Athey, Susan , title =. Journal of the American Statistical Association , year =
[3]

The Annals of Applied Statistics , year =

Meng, Xiao-Li , title =. The Annals of Applied Statistics , year =

work page
[4]

Elements of Causal Inference: Foundations and Learning Algorithms , publisher =

Peters, Jonas and Janzing, Dominik and Sch. Elements of Causal Inference: Foundations and Learning Algorithms , publisher =. 2017 , address =

work page 2017
[5]

Causal Inference by Using Invariant Prediction: Identification and Confidence Intervals , journal =

Peters, Jonas and B. Causal Inference by Using Invariant Prediction: Identification and Confidence Intervals , journal =. 2016 , volume =

work page 2016
[6]

, title =

Hill, Jennifer L. , title =. Journal of Computational and Graphical Statistics , year =

work page
[7]

, title =

LaLonde, Robert J. , title =. The American Economic Review , year =

work page
[8]

2018 , publisher=

Double/debiased machine learning for treatment and structural parameters , author=. 2018 , publisher=

2018

Show all 18 references

[9]

2005 , publisher=

Algorithmic learning in a random world , author=. 2005 , publisher=

2005
[10]

Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

Conformal inference of counterfactuals and individual treatment effects , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2021 , publisher=

2021
[11]

arXiv preprint arXiv:2006.01474 , year=

Conformal prediction intervals for the individual treatment effect , author=. arXiv preprint arXiv:2006.01474 , year=

2006 arXiv
[12]

Proceedings of the national academy of sciences , volume=

Metalearners for estimating heterogeneous treatment effects using machine learning , author=. Proceedings of the national academy of sciences , volume=. 2019 , publisher=

2019
[13]

Biometrika , volume=

Quasi-oracle estimation of heterogeneous treatment effects , author=. Biometrika , volume=. 2021 , publisher=

2021
[14]

Econometrica: journal of the Econometric Society , pages=

Root-N-consistent semiparametric regression , author=. Econometrica: journal of the Econometric Society , pages=. 1988 , publisher=

1988
[15]

Journal of the American statistical Association , volume=

Estimation of regression coefficients when some regressors are not always observed , author=. Journal of the American statistical Association , volume=. 1994 , publisher=

1994
[16]

Behavioral predictive modeling in economics , pages=

Using machine learning methods to support causal inference in econometrics , author=. Behavioral predictive modeling in economics , pages=. 2020 , publisher=

2020
[17]

2015 , publisher=

Causal inference in statistics, social, and biomedical sciences , author=. 2015 , publisher=

2015
[18]

arXiv preprint arXiv:2605.25687 , year=

Confidence intervals for causal effects in sequential decision making , author=. arXiv preprint arXiv:2605.25687 , year=

arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

2026 , month = may, howpublished =

Cunningham, Scott , title =. 2026 , month = may, howpublished =

work page 2026

[2] [2]

Journal of the American Statistical Association , year =

Wager, Stefan and Athey, Susan , title =. Journal of the American Statistical Association , year =

[3] [3]

The Annals of Applied Statistics , year =

Meng, Xiao-Li , title =. The Annals of Applied Statistics , year =

work page

[4] [4]

Elements of Causal Inference: Foundations and Learning Algorithms , publisher =

Peters, Jonas and Janzing, Dominik and Sch. Elements of Causal Inference: Foundations and Learning Algorithms , publisher =. 2017 , address =

work page 2017

[5] [5]

Causal Inference by Using Invariant Prediction: Identification and Confidence Intervals , journal =

Peters, Jonas and B. Causal Inference by Using Invariant Prediction: Identification and Confidence Intervals , journal =. 2016 , volume =

work page 2016

[6] [6]

, title =

Hill, Jennifer L. , title =. Journal of Computational and Graphical Statistics , year =

work page

[7] [7]

, title =

LaLonde, Robert J. , title =. The American Economic Review , year =

work page

[8] [8]

2018 , publisher=

Double/debiased machine learning for treatment and structural parameters , author=. 2018 , publisher=

2018