Pith. sign in

REVIEW 6 major objections 5 minor 50 references

The optimal estimator for an RCT is a property of the trial family and the analytical goal, not of the estimator alone; a cross-fitting framework identifies it using real historical trials.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-07-31 23:56 UTC pith:UQ2IIXTR

load-bearing objection A correct but narrow cross-fitted estimator comparison framework; the empirical recommendations are in-sample argmins and the abstract contradicts the main results—worth peer review, not acceptance as-is. the 6 major comments →

arxiv 2607.23254 v1 pith:UQ2IIXTR submitted 2026-07-25 stat.AP econ.EMstat.ME

Towards Optimal Estimators for Randomized Control Trials

Harsh Parikh , Gabriel Levin-Konigsberg , Nilesh Tripuraneni , Dhruv Madeka , Michael I. Jordan , Dean Foster , Dominique Perrault-Joncas , Alexander Volfovsky This is my paper
classification stat.AP econ.EMstat.ME
keywords randomized controlled trialsestimator selectioncross-fittingmean squared errorregretcovariate adjustmentwinsorizationcausal inference
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

No single estimator is best for every experiment. The paper argues that the optimal estimator for a treatment effect is tied to a family of related randomized trials and to the analyst's goal: minimizing mean squared error for inference or minimizing regret for rollout decisions. To find that estimator without knowing true effects, it proposes a cross-fitting procedure in which a difference-in-means estimate on one data fold serves as a noisy benchmark for any estimator trained on the other fold; a theorem shows the resulting MSE comparison is unbiased, while the regret comparison has bias that shrinks as studies grow. Applied to hundreds of real trials, the framework finds that simple difference-in-means (no winsorization) minimizes regret, while covariate-adjusted, winsorized estimators (e.g., Linear-T_0.005) minimize MSE. The upshot is actionable: decision-makers and inference-focused researchers should deliberately choose different estimators.

Core claim

The central claim is that no estimator is universally optimal; the best estimator belongs to a family of related experiments and an evaluation metric. Given studies that share an outcome and satisfy covariate-conditional study independence, the paper defines expected MSE and expected regret for each estimator and constructs a cross-fitted test statistic: split each study into K folds, fit the candidate estimator on K−1 folds, and score it against a difference-in-means estimate on the held-out fold, which is unbiased for the true effect. Theorem 4.1 proves the MSE comparison is unbiased; the regret comparison has bias bounded by interpretable sign-error terms and vanishes asymptotically. Empi

What carries the argument

Cross-fitted performance comparison on real trial families. For each study the sample is split into K folds; the candidate estimator is trained on the complement and evaluated against a difference-in-means estimate from the held-out fold, using squared error (MSE) or a sign-based regret that captures the cost of wrong launch decisions. The paper's Theorem 4.1 shows the MSE statistic is unbiased, and Theorem 4.2 gives a bias bound for regret that vanishes asymptotically, turning collections of historical RCTs into a testing ground for estimator rankings.

Load-bearing premise

Assumption A.2: conditional on measured covariates, potential outcomes are independent of which study a unit belongs to; if unmeasured study-level differences affect outcomes, the best estimator on past trials may not be best on the next one.

What would settle it

Split a family of historical trials into two non-overlapping halves; use the framework to pick the top estimator on the first half, then measure its MSE and regret on the second half. If the top-ranked estimator is not top-ranked on the held-out half, the transfer assumption fails. A more direct test would show that study membership predicts potential outcomes after adjusting for all observed covariates, e.g., via a placebo test on the trial design.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Organizations running many trials on the same outcome can pre-register the estimator the framework ranks first, replacing post-hoc estimator choice with a data-driven rule.
  • Estimator rankings can reverse between MSE and regret, so a single 'best' estimator for both inference and decisions is provably suboptimal on at least one objective.
  • The unbiasedness of the cross-fitted MSE statistic means performance gaps observed in the case studies reflect real differences, not overfitting to individual trials.
  • Regret comparisons carry a conservative finite-sample bias: sign errors in the benchmark difference-in-means make regret look smaller, a distortion that disappears only as study sizes grow.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The framework doubles as a portfolio-level audit: an organization could rerun it periodically to check whether its default estimator still matches its stated objective as trial populations drift.
  • A natural extension is to weaken the covariate-conditional study independence assumption with a hierarchical or sensitivity model that allows mechanism differences across studies; without such an extension, rankings are only guaranteed within a tightly controlled family.
  • A concrete finite-sample question the paper leaves open is how many studies and how many units per study are needed before the ranking is stable; simulation or resampling from the case-study data could answer it.
  • The regret metric could be generalized to asymmetric costs of false positives vs. false negatives; the DM_0 dominance found here may not survive when misses are weighted much more heavily than false alarms.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

6 major / 5 minor

Summary. The paper proposes a framework for choosing treatment-effect estimators within families of RCTs. It defines MSE and regret performance metrics, estimates them via within-study cross-fitting using the difference-in-means estimate on a held-out fold as a proxy for the truth, and builds a permutation test for comparing estimators. The method is applied to 556 Amazon Supply Chain Optimization Technology trials and to 25 Strengthening Democracy Challenge interventions. The paper claims that the optimal estimator is objective-dependent: a covariate-adjusted, winsorized estimator minimizes MSE for inference, while unadjusted difference-in-means minimizes regret for decision-making. Theoretical results include an unbiasedness theorem for the cross-fitted MSE test statistic and bias bounds for the regret-based statistic.

Significance. If the framework is valid, it is a useful step toward principled, domain-specific estimator selection, exploiting the growing availability of large portfolios of real experiments. The use of within-study cross-fitting to separate the candidate estimator from the benchmark is sound for MSE comparisons, and the 556-trial Amazon application is unusually rich. However, the paper's central theoretical claim is not established as stated, the regret comparisons have an acknowledged but unquantified bias that can favor the benchmark estimator, and the key cross-study generalization assumption is untested. The empirical recommendations are also internally inconsistent (WLS vs Linear-T in the abstract/body). The core idea is promising, but the current version does not yet support the headline claims.

major comments (6)
  1. [Section 4.3 and Appendix B, Eqs. (5)-(8)] Theorem 4.1 is not established as stated because the proof equates the expectation of the cross-fitted squared error with the full-sample performance. Eq. (5) defines epsilon_alpha^(s) using tau_hat_alpha^(s) on the full sample O_s, but the cross-fitted estimate uses tau_hat_(alpha,-k) fitted on O_(s,-k), which has size (1-1/K)m_s. MSE depends on sample size, so E[(tau_hat_(alpha,-k)-tau)^2 | S=s] differs from epsilon_alpha^(s) in finite samples. Unbiasedness would require redefining epsilon_alpha as the split-sample estimator's performance; then the optimal estimator selected by the framework is not the full-sample estimator recommended in Section 5.2.
  2. [Section 3.1, Assumption A.2] Assumption A.2, which the paper calls 'crucial,' licenses the transition from 'best on the observed 556 studies' to 'best for the family.' The Amazon studies differ by geography, marketplace, product category, and treatment, and the covariates X_i are not described in enough detail to support the claim that they capture all mechanism-relevant differences. Because most treatments are unique, A.2 is largely unfalsifiable from these data. If A.2 fails, the pooled optimum need not generalize to the next trial. The theory simply assumes A.2; the conclusions should be framed as conditional on this assumption or the assumption should be relaxed/tested.
  3. [Abstract vs. Section 5.2] The abstract states that 'weighted least squares performs best for inference goals,' but Section 5.2 reports that Linear-T_0.005 achieves minimum MSE on the Amazon data, with OLS optimal at winsorization levels below 0.5%. No WLS result is identified as optimal in the body. This contradiction in the paper's headline empirical claim must be resolved before publication.
  4. [Section 4.3 and Appendix B, Theorem 4.3] For the regret metric, Theorem 4.3 gives E[hat_theta_regret - theta_regret] = (1/N) sum_s (B_s^(beta) - B_s^(alpha)), which is not zero. The bias is not corrected or bounded empirically. Because the difference-in-means estimate hat_tau_DM,k serves as the surrogate truth, the DM estimator's own decision rule is correlated with the benchmark, potentially giving DM_0 a systematic advantage in the regret comparison. The finding that 'DM_0 minimizes regret' may be an artifact of this bias; the paper should quantify the bias or temper the conclusion.
  5. [Section 4.4, permutation test] The permutation test is justified by claiming that H0: theta(alpha,beta)=0 implies exchangeability of estimator labels on the per-study performance estimates. This is false: equal expected performance does not imply that the two estimators' performance scores are exchangeable. A permutation test of label exchangeability will reject when the estimators have equal means but different variances. A valid test for equality of means, such as a bootstrap or studentized permutation test, is needed.
  6. [Section 5.1, multi-arm trials] The paper says that for multi-arm trials it 'construct[s] pairwise comparisons between each treatment and control.' If each treatment-control pair is treated as an independent study, then control arms are shared across observations, violating Assumption A.3 (non-overlapping study samples) and inducing dependence in the 556-study aggregate. The analysis should either use the original trial as the unit or account for clustering within multi-arm trials.
minor comments (5)
  1. [Section 1] Typo: 'the the standard difference-in-means estimator'.
  2. [Section 7] Typo: 'minimizing regreet' should be 'minimizing regret'.
  3. [Section A.1] The SDC outcome description says SPV is measured through 'three 7-point scale items' but only SPV 1 and SPV 2 are listed; also Figure 3's caption says 'six estimators' while the legend shows four.
  4. [Section 4.5] The weighted extension estimates w^(s) on the full data. This breaks the independence between the weights, the candidate estimator, and the DM benchmark used in Theorem 4.1. No theoretical guarantee is given for the weighted test statistic.
  5. [References] Some references are incomplete or appear mismatched: Parikh et al. (2022) has no venue/arXiv number, and the Losch et al. (2021) citation title concerns optimal transport rather than a comparison of synthetic versus real-world estimator rankings.

Circularity Check

0 steps flagged

No significant circularity: cross-fitted comparisons are genuine paired tests; empirical recommendations are in-sample summaries, not predictions.

full rationale

The paper's central derivation (Theorems 4.1-4.3) is self-contained. The cross-fitted MSE statistic compares each estimator against a common difference-in-means reference computed on an independent fold; the DM error term cancels in pairwise differences, and the cross term vanishes by sample-splitting independence. This is a substantive paired-comparison argument, not a definitional identity. The regret bounds in Theorems 4.2-4.3 are derived explicitly from the regret definition rather than assumed. The empirical findings (e.g., DM_0 for regret, Linear-T_0.005 for MSE in the Amazon body; OLS_0.1 for some SDC outcomes) are in-sample minimizers of the estimated metrics over the observed 556/25 studies; the paper does not present them as out-of-sample predictions, and its 'actionable guidance' is explicitly conditioned on the observed family. The step from observed best to family-optimal is mediated by Assumption A.2, which is a substantive, potentially untestable premise about cross-study exchangeability, not a circular definition. The only overlapping-author citation (Parikh et al. 2022) is used as context on synthetic-data limitations and is not load-bearing. No estimator is 'predicted' from a parameter fitted to the same target, and no uniqueness theorem or ansatz is imported from the authors' prior work. A separate internal inconsistency (abstract says WLS best for inference while the body says Linear-T_0.005) is a correctness/consistency concern, not circularity.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 0 invented entities

No new physical or statistical entities are invented. The uncharged assumptions are all statistical: the strong A.2 cross-study invariance, the use of DM as a noisy benchmark, and a menu of user-selected tuning parameters (winsorization grid, WLS weights, inverse-variance weights, c, K). These choices are not fitted to derive the theorems, but the headline empirical optimality claims do depend on them.

free parameters (5)
  • Winsorization grid = p ∈ {0, 0.001, 0.005, 0.01, 0.025, 0.05, 0.1}
    The set of winsorization levels defines the 28 estimator variants; the reported 'optimal' estimator depends on this arbitrary grid (e.g., Linear-T_0.005 best at Amazon, OLS best below 0.5%).
  • WLS weight function = w_i = 1/(Y_pre,i − \bar Y_pre)^2
    Ad hoc inverse-distance weighting used in the WLS estimator; no justification is given, and it affects the Amazon rankings.
  • Inverse-variance study weights = w(s) = 1 / Var(\hat τ_DM(s))
    Estimated from the same data used for evaluation in Section 4.5; the weighted extension changes M_α and can change rankings, but no weighted ranking results are reported.
  • Critical threshold c and regret normalization = c ∈ R_+; regret normalized at c=1.96 for DM_0
    Regret curves and 'minimum regret' rankings depend on the chosen threshold range and normalization; DM_0 is reported best only relative to the chosen c grid.
  • Number of cross-fitting folds K = unspecified
    K is never stated explicitly; unbiasedness holds for any fixed K but finite-sample variance and the regret bias bounds depend on it.
axioms (6)
  • domain assumption A.1 Randomization within studies: Yi(t) ⟂ Ti | S=s, with overlap
    Standard RCT assumption used to identify τ(s) and to make the difference-in-means estimator unbiased for the true effect (Section 3.1).
  • domain assumption A.2 Covariate-Conditional Study Independence: Yi(t) ⟂ S | Xi
    Crucial for pooling trials into a family and for transferring rankings to future studies; unverifiable in the Amazon SCOT data where studies differ by geography, marketplace, and product category.
  • domain assumption A.3 Non-overlapping study samples: Os ∩ Os' = ∅
    Ensures independence across studies; likely reasonable in the case studies but not guaranteed in general experimentation platforms.
  • domain assumption Superpopulation two-stage sampling of studies and treatments
    The framework assumes studies are draws from a population, enabling M_α = E_S[ε(s)_α]; without this, the 'family' is just the observed set.
  • domain assumption Exchangeability of estimator labels under H0 for permutation inference
    Permutation p-values in Section 4.4 require exchangeability of performance differences across studies under equal performance; this is asserted, not proved.
  • domain assumption DM consistency / sign consistency for asymptotic unbiasedness of regret comparisons
    Corollary 4.1 relies on DM converging to τ(s); for heavy-tailed Amazon outcomes with finite samples, this is an asymptotic hope rather than a finite-sample guarantee.

pith-pipeline@v1.3.0-alltime-deepseek · 14632 in / 14754 out tokens · 132841 ms · 2026-07-31T23:56:23.760660+00:00 · methodology

0 comments
read the original abstract

Randomized controlled trials (RCTs) are fundamental tools for causal inference across technology companies, pharmaceutical research, and federal agencies. While the standard difference-in-means estimator provides unbiased treatment effect estimates, it often lacks precision, particularly when treatment effects are heterogeneous or outcomes exhibit heavy-tailed distributions. Although numerous precision-enhancing methods exist---from covariate adjustment techniques to variance reduction strategies---recent research demonstrates that no single estimator performs optimally across all datasets. Rather than seeking the best estimator for individual RCTs, which risks compromising scientific validity through convenient selection, we propose a principled framework for identifying optimal estimators within families of RCTs based on specific analytical goals. Our approach uses sample splitting to estimate the distribution of evaluation metrics (e.g., mean squared error, regret) across RCT families, enabling systematic comparisons between estimators while maintaining asymptotic guarantees. We demonstrate this framework using a sample of Amazon's Supply Chain Optimization Technology trials and the Strengthening Democracy Challenge dataset (25 interventions). Results reveal that optimal estimators vary significantly by analytical objective: weighted least squares performs best for inference goals, while difference-in-means minimizes regret for decision-making contexts. This work provides actionable guidance for estimator selection while preserving methodological rigor across diverse research applications.

Figures

Figures reproduced from arXiv: 2607.23254 by Alexander Volfovsky, Dean Foster, Dhruv Madeka, Dominique Perrault-Joncas, Gabriel Levin-Konigsberg, Harsh Parikh, Michael I. Jordan, Nilesh Tripuraneni.

Figure 1
Figure 1. Figure 1: Mean Squared Error performance across treatment effect estimators in Amazon [PITH_FULL_IMAGE:figures/full_fig_p019_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (a) Minimum “regret” performance across treatment effect estimators in Amazon [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Mean squared error and standard deviation for six estimators across outcomes. [PITH_FULL_IMAGE:figures/full_fig_p023_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Regret Curves estimators as a function of critical-t across outcomes. Lower values [PITH_FULL_IMAGE:figures/full_fig_p024_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Minimum regret achieved by each estimator across outcomes. [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 4 linked inside Pith

  1. [1]

    2008 , publisher=

    Mostly Harmless Econometrics , author=. 2008 , publisher=

    2008

  2. [2]

    Proceedings of the National Academy of Sciences , volume=

    Recursive partitioning for heterogeneous causal effects , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

    2016

  3. [3]

    Handbook of Economic Field Experiments , volume=

    The Econometrics of Randomized Experiments , author=. Handbook of Economic Field Experiments , volume=. 2017 , publisher=

    2017

  4. [4]

    Athey, Susan and Imbens, Guido and Metzger, Jonas and Munro, Evan , year=. Using. 1909.02210 , archivePrefix=

    Pith/arXiv arXiv 1909

  5. [5]

    The Annals of Statistics , pages=

    Bootstrap of the mean in the infinite variance case , author=. The Annals of Statistics , pages=. 1987 , publisher=

    1987

  6. [6]

    arXiv preprint arXiv:2104.00673 , year=

    Cross-validation: what does it estimate and how well does it do it? , author=. arXiv preprint arXiv:2104.00673 , year=

    Pith/arXiv arXiv

  7. [7]

    2020 , journal=

    Cross-validation confidence intervals for test error , author=. 2020 , journal=

    2020

  8. [8]

    Journal of Machine Learning Research , volume=

    No unbiased estimator of the variance of k-fold cross-validation , author=. Journal of Machine Learning Research , volume=

  9. [9]

    Improving precision and power in randomized trials for

    Benkeser, David and D. Improving precision and power in randomized trials for. 2020 , month = oct, publisher =. doi:10.1111/biom.13377 , url =

    doi:10.1111/biom.13377 2020

  10. [10]

    2022 , publisher=

    Covariate Adjustment in Randomized Trials , author=. 2022 , publisher=

    2022

  11. [11]

    Proceedings of the National Academy of Sciences , volume=

    Lasso Adjustments of Treatment Effect Estimates in Randomized Experiments , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

    2016

  12. [12]

    2020 , journal=

    Optimal mean estimation without a variance , author=. 2020 , journal=

    2020

  13. [13]

    2018 , publisher=

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

    2018

  14. [14]

    2023 , eprint=

    A First Course in Causal Inference , author=. 2023 , eprint=

    2023

  15. [15]

    How to make a

    Drees, Holger and Resnick, Sidney and de Haan, Laurens , journal=. How to make a. 2000 , publisher=

    2000

  16. [16]

    1998 , institution=

    Statistical Principles for Clinical Trials (. 1998 , institution=

    1998

  17. [17]

    2014 , journal=

    Semiparametric Exponential Families for Heavy-Tailed Data , author=. 2014 , journal=

    2014

  18. [18]

    Advances in Applied Mathematics , volume=

    On Regression Adjustments to Experimental Data , author=. Advances in Applied Mathematics , volume=. 2008 , publisher=

    2008

  19. [19]

    American Journal of Epidemiology , volume=

    Doubly Robust Estimation of Causal Effects , author=. American Journal of Epidemiology , volume=. 2011 , publisher=

    2011

  20. [20]

    Journal of the American Statistical Association , volume=

    The predictive sample reuse method with applications , author=. Journal of the American Statistical Association , volume=. 1975 , publisher=

    1975

  21. [21]

    Advances in Neural Information Processing Systems , volume=

    The Case for Evaluating Causal Models Using Interventional Measures and Empirical Data , author=. Advances in Neural Information Processing Systems , volume=. 2019 , url=

    2019

  22. [22]

    Hadad, Vitor , year=

  23. [23]

    2020 , publisher=

    Causal Inference: What If , author=. 2020 , publisher=

    2020

  24. [24]

    2015 , publisher=

    Causal Inference in Statistics, Social, and Biomedical Sciences , author=. 2015 , publisher=

    2015

  25. [25]

    and Sollecito, William A

    Koch, Gary G. and Sollecito, William A. , title =. Drug Information Journal , volume =. 1984 , doi =

    1984

  26. [26]

    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

  27. [27]

    Journal of the American Statistical Association , volume=

    Cross-validation with confidence , author=. Journal of the American Statistical Association , volume=. 2020 , publisher=

    2020

  28. [28]

    Agnostic Notes on Regression Adjustments to Experimental Data: Reexamining

    Lin, Winston , journal=. Agnostic Notes on Regression Adjustments to Experimental Data: Reexamining. 2013 , publisher=

    2013

  29. [29]

    arXiv preprint arXiv:2112.09266 , year=

    Optimal Transport of Data Generates Distributional Robustness , author=. arXiv preprint arXiv:2112.09266 , year=

    Pith/arXiv arXiv

  30. [30]

    Foundations of Computational Mathematics , volume=

    Mean estimation and regression under heavy-tailed distributions: A survey , author=. Foundations of Computational Mathematics , volume=. 2019 , publisher=

    2019

  31. [31]

    Essay on principles

    On the application of probability theory to agricultural experiments. Essay on principles. Section 9 , author=. Statistical Science , pages=. 1990 , publisher=

    1990

  32. [32]

    , author=

    Estimating causal effects of treatments in randomized and nonrandomized studies. , author=. Journal of educational Psychology , volume=. 1974 , publisher=

    1974

  33. [33]

    International Conference on Machine Learning , pages=

    Orthogonal random forest for causal inference , author=. International Conference on Machine Learning , pages=. 2019 , organization=

    2019

  34. [34]

    2022 , eprint=

    Validating Causal Inference Methods , author=. 2022 , eprint=

    2022

  35. [35]

    Statistics in Medicine , volume=

    Some methods for heterogeneous treatment effect estimation in high dimensions , author=. Statistics in Medicine , volume=. 2018 , publisher=

    2018

  36. [36]

    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

  37. [37]

    Journal of the American Statistical Association , volume=

    Causal inference using potential outcomes: Design, modeling, decisions , author=. Journal of the American Statistical Association , volume=. 2005 , publisher=

    2005

  38. [38]

    and Merlin, Vincent R

    Saari, Donald G. and Merlin, Vincent R. , journal=. The. 1996 , publisher=

    1996

  39. [39]

    arXiv preprint arXiv:1804.05146 , year=

    A comparison of methods for model selection when estimating individual treatment effects , author=. arXiv preprint arXiv:1804.05146 , year=

    Pith/arXiv arXiv

  40. [40]

    Journal of the Royal Statistical Society: Series B (Methodological) , volume=

    Cross-validatory choice and assessment of statistical predictions , author=. Journal of the Royal Statistical Society: Series B (Methodological) , volume=. 1974 , publisher=

    1974

  41. [41]

    Conference on Learning Theory , pages=

    Estimation and inference with trees and forests in high dimensions , author=. Conference on Learning Theory , pages=. 2020 , organization=

    2020

  42. [42]

    2016 , journal=

    Scalable semiparametric inference for the means of heavy-tailed distributions , author=. 2016 , journal=

    2016

  43. [43]

    Statistics in Medicine , volume=

    Covariate Adjustment for Two-Sample Treatment Comparisons in Randomized Clinical Trials: A Principled yet Flexible Approach , author=. Statistics in Medicine , volume=. 2008 , publisher=

    2008

  44. [44]

    and Stagnaro, Michael N

    Voelkel, Jan G. and Stagnaro, Michael N. and Chu, James Y. and Pink, Sophia L. and Mernyk, Joseph S. and Redekopp, Chrystal and Ghezae, Isaias and Cashman, Matthew and Adjodah, Dhaval and Allen, Levi G. and Allis, L. Victor and Baleria, Gina and Ballantyne, Nathan and Van Bavel, Jay J. and Blunden, Hayley and Braley, Alia and Bryan, Christopher J. and Cel...

    2024

  45. [45]

    Proceedings of the National Academy of Sciences , volume=

    High-Dimensional Regression Adjustments in Randomized Experiments , author=. Proceedings of the National Academy of Sciences , volume=. 2016 , publisher=

    2016

  46. [46]

    Econometrica , volume=

    Efficient estimation of average treatment effects using the estimated propensity score , author=. Econometrica , volume=. 2003 , publisher=

    2003

  47. [47]

    2018 , journal=

    Estimation and Inference of Heterogeneous Treatment Effects using Random Forests , author=. 2018 , journal=

    2018

  48. [48]

    Wager, Stefan , year=

  49. [49]

    Econometrica: Journal of the Econometric Society , pages=

    A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity , author=. Econometrica: Journal of the Econometric Society , pages=. 1980 , publisher=

    1980

  50. [50]

    The American Statistician , volume=

    Efficiency Study of Estimators for a Treatment Effect in a Pretest-Posttest Trial , author=. The American Statistician , volume=. 2001 , publisher=

    2001