pith. machine review for the scientific record. sign in

arxiv: 2605.02154 · v2 · submitted 2026-05-04 · 📊 stat.ME

Recognition: 2 theorem links

· Lean Theorem

Efficient Transported Distributional and Quantile Treatment Effects with Surrogate-Assisted Missing Primary Outcomes

Pengyun Wang

Pith reviewed 2026-05-08 19:25 UTC · model grok-4.3

classification 📊 stat.ME
keywords transported treatment effectsquantile treatment effectsdistributional treatment effectssurrogate outcomesmissing at randomefficient influence functionone-step estimatorcausal inference
0
0 comments X

The pith

Surrogates improve efficiency for estimating transported distributional and quantile treatment effects when primary outcomes are missing at random.

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

The paper focuses on estimating how treatment affects the full distribution and quantiles of a long-run primary outcome in a target population. Source data includes treatment assignment, post-treatment surrogates for everyone, and the primary outcome only for a validation subset, while the target population supplies only baseline covariates. Surrogates are not assumed to perfectly replace the primary outcome. Instead, they enter the estimation to reduce variance under missing-at-random sampling of the primary outcome. The central advance is the derivation of a nonparametric efficient influence function whose three orthogonal components separately handle target covariate sampling, the source surrogate process, and the missing primary outcomes, which then produces a simple cross-fitted one-step estimator.

Core claim

The nonparametric efficient influence function for the transported counterfactual distribution functions, their quantiles, and the quantile treatment effect decomposes into three orthogonal components that correspond to target covariate sampling, the source surrogate process, and missing primary outcomes. This decomposition yields a closed-form cross-fitted one-step estimator after nuisance estimation and delivers pointwise and uniform asymptotic linearity for the distribution functions, Bahadur representations for the quantiles, multiplier-bootstrap simultaneous bands, and explicit efficiency gains from the surrogates.

What carries the argument

The nonparametric efficient influence function with three orthogonal components for transported counterfactual distributions and quantiles.

If this is right

  • The transported counterfactual distribution functions and quantiles are identified under the stated missing-at-random and transportability conditions.
  • The estimator for the distribution functions is asymptotically linear and the quantile estimator admits a Bahadur representation under explicit local inverse-map conditions.
  • Multiplier-bootstrap simultaneous confidence bands are valid under stated conditions on the estimated process and densities.
  • Observing surrogates produces measurable efficiency gains relative to using only the validation subset of primary outcomes.
  • The nuisance estimators can be implemented with sieve ridge regression, kernel ridge regression, or isotonic projection while satisfying the required rate conditions.

Where Pith is reading between the lines

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

  • The same three-component influence-function strategy could be applied to other causal estimands that combine transport and missing data.
  • In applications such as long-term follow-up studies, the efficiency gain from surrogates may allow smaller validation samples while preserving precision.
  • The approach separates the roles of surrogates from any assumption that they are perfect substitutes, which may encourage their use in settings where surrogacy is implausible.

Load-bearing premise

Primary outcomes are missing at random conditional on the observed data that includes surrogates, together with transportability assumptions that let the source study inform the target population.

What would settle it

A Monte Carlo experiment or real-data example in which the proposed one-step estimator fails to achieve the variance lower bound implied by the derived influence function, even when the nuisance functions are estimated at the required rates.

Figures

Figures reproduced from arXiv: 2605.02154 by Pengyun Wang.

Figure 1
Figure 1. Figure 1: ACTG 175 empirical illustration: transported QTE curve for 96-week CD4 count. view at source ↗
read the original abstract

We study target-population distributional and quantile treatment effects when a source study observes treatment and post-treatment surrogates for all source units but observes a long-run primary outcome only for a validation subset, while the target population contributes only baseline covariates. The target estimands are transported counterfactual distribution functions $\psi_a(y)=P(Y^a\le y\mid R=0)$, their quantiles $q_a(\tau)$, and the quantile treatment effect $\Delta(\tau)=q_1(\tau)-q_0(\tau)$. The surrogate is not treated as a replacement endpoint and no Prentice-type surrogacy condition is imposed. Instead, the surrogate is used only to improve efficiency under missing-at-random primary-outcome sampling. We derive the nonparametric efficient influence function, which has three orthogonal components corresponding to target covariate sampling, the source surrogate process, and missing primary outcomes. This yields a closed-form cross-fitted one-step estimator after nuisance estimation. We establish identification, the canonical gradient, exact drift identities, ratio-level robustness, pointwise and uniform asymptotic linearity for transported CDFs, Bahadur representations for quantiles under explicit local inverse-map conditions, high-level multiplier-bootstrap simultaneous bands under explicit estimated-process and density conditions, and quantile-specific efficiency gains from observing surrogates. We also give lower-level nuisance-rate verification for a deliberately restricted class of analyzable bounded finite-dimensional or finite-rank implementations based on sieve ridge regression, ridge logistic regression, calibrated density-ratio estimation, finite-rank kernel ridge regression, and isotonic projection under explicit grid, eigenvalue, source, and entropy conditions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 4 minor

Summary. The paper studies estimation of transported counterfactual distribution functions ψ_a(y) = P(Y^a ≤ y | R=0), their quantiles q_a(τ), and the quantile treatment effect Δ(τ) when source data include treatment, surrogates for all units, and primary outcomes only for a validation subset, while target data supply only baseline covariates. It derives the nonparametric efficient influence function as the sum of three orthogonal components (target covariate sampling, source surrogate process, and MAR missingness for primaries), constructs a closed-form cross-fitted one-step estimator after nuisance estimation, and proves identification, asymptotic linearity, Bahadur representations for quantiles, multiplier-bootstrap validity, and efficiency gains, together with lower-level rate conditions for sieve/ridge/kernel implementations.

Significance. If the derivations hold, the work supplies a principled, efficiency-improving approach to transported distributional and quantile effects that uses surrogates solely for precision under MAR without invoking Prentice surrogacy. Strengths include the explicit three-component EIF, exact drift identities, ratio-level robustness, uniform asymptotic results, and concrete nuisance-rate verifications for bounded finite-rank estimators. These features address a practically relevant design with partial outcome observation and population transport.

minor comments (4)
  1. [§2.2] §2.2, Assumption 3 (MAR): the conditioning set is described as 'observed data including surrogates,' but the precise sigma-field (e.g., whether it includes the treatment indicator A) should be stated explicitly to match the EIF derivation in §3.1.
  2. [Theorem 4.1] Theorem 4.1 (asymptotic linearity): the local inverse-map condition for quantiles is stated at a high level; a brief remark on how the density lower bound is verified for the isotonic projection implementation would aid reproducibility.
  3. [Figure 2] Figure 2 (simulation results): the legend for the 'no-surrogate' comparator is missing; adding it would clarify the efficiency-gain comparison reported in the text.
  4. [§5.3] §5.3, nuisance estimation: the entropy condition for the kernel ridge class is given, but the explicit constant in the eigenvalue decay rate (used for the rate verification) is not numerically illustrated; a short table entry would help.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and accurate summary of our manuscript on efficient transported distributional and quantile treatment effects with surrogate-assisted missing primary outcomes. The recommendation for minor revision is appreciated, and we note that the referee correctly identifies the key contributions including the three-component EIF, cross-fitted one-step estimator, and asymptotic results. As no specific major comments were raised, we have no point-by-point rebuttals to provide at this stage.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The central derivation derives the nonparametric EIF explicitly from the observed-data model under transportability and MAR assumptions conditional on surrogates and covariates, with three orthogonal components obtained via pathwise differentiability. The one-step estimator is constructed directly from this EIF after cross-fitting nuisance functions; no target estimand is redefined as a fitted parameter, no uniqueness theorem is imported from self-citations, and no ansatz is smuggled via prior work. All steps (identification, canonical gradient, asymptotic linearity, Bahadur representations) are self-contained against the stated high-level conditions and do not reduce by construction to the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper rests on standard causal identification assumptions for transportability and missing-at-random sampling; no free parameters or new entities are introduced, as the approach is nonparametric and derives the EIF directly.

axioms (2)
  • domain assumption Missing at random for primary outcomes conditional on observed data including surrogates
    Invoked to allow the surrogate to improve efficiency without replacing the primary outcome.
  • domain assumption Standard transportability identification assumptions linking source and target populations
    Required to identify the target counterfactual distributions from the source data.

pith-pipeline@v0.9.0 · 5579 in / 1454 out tokens · 23778 ms · 2026-05-08T19:25:11.036009+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Nested Sensitivity Envelopes for Transported Quantile Treatment Effects

    stat.ME 2026-05 unverdicted novelty 7.0

    Derives sharp nested CDF envelopes for transported quantile treatment effects under marginal sensitivity bounds on confounding and transportability, with semiparametric estimators, uniform inference, and breakdown frontiers.

Reference graph

Works this paper leans on

53 extracted references · 4 canonical work pages · cited by 1 Pith paper

  1. [1]

    W., and Kang, H

    Athey, S., Chetty, R., Imbens, G. W., and Kang, H. (2025). The surrogate index: Combining short-term proxies to estimate long-term treatment effects more rapidly and precisely. The Review of Economic Studies, rdaf087

    2025

  2. [2]

    and Robins, J

    Bang, H. and Robins, J. M. (2005). Doubly robust estimation in missing data and causal inference models. Biometrics, 61:962--973

    2005

  3. [3]

    Boughdiri, A., Berenfeld, C., Josse, J., and Scornet, E. (2025). A unified framework for the transportability of population-level causal measures. arXiv:2505.13104

    work page arXiv 2025

  4. [4]

    and Molenberghs, G

    Buyse, M. and Molenberghs, G. (1998). Criteria for the validation of surrogate endpoints in randomized experiments. Biometrics, 54:1014--1029

    1998

  5. [5]

    and Cai, T

    Chakrabortty, A. and Cai, T. (2018). Efficient and adaptive linear regression in semi-supervised settings. The Annals of Statistics, 46:1541--1572

    2018

  6. [6]

    Chakrabortty, A., Dai, G., and Carroll, R. J. (2022). Semi-supervised quantile estimation: Robust and efficient inference in high dimensional settings. arXiv:2201.10208

    work page arXiv 2022

  7. [7]

    and Dai, G

    Chakrabortty, A. and Dai, G. (2024). A general framework for treatment effect estimation in semi-supervised and high dimensional settings. arXiv:2201.00468

    work page arXiv 2024

  8. [8]

    and Ritzwoller, D

    Chen, J. and Ritzwoller, D. M. (2023). Semiparametric estimation of long-term treatment effects. Journal of Econometrics, 237:105545

    2023

  9. [9]

    Cole, S. R. and Stuart, E. A. (2010). Generalizing evidence from randomized clinical trials to target populations: The ACTG 320 trial. American Journal of Epidemiology, 172:107--115

    2010

  10. [10]

    J., Robertson, S

    Dahabreh, I. J., Robertson, S. E., Tchetgen Tchetgen, E. J., Stuart, E. A., and Hern\'an, M. A. (2019). Generalizing causal inferences from individuals in randomized trials to all trial-eligible individuals. Biometrics, 75:685--694

    2019

  11. [11]

    and S\"arndal, C.-E

    Deville, J.-C. and S\"arndal, C.-E. (1992). Calibration estimators in survey sampling. Journal of the American Statistical Association, 87:376--382

    1992

  12. [12]

    Frangakis, C. E. and Rubin, D. B. (2002). Principal stratification in causal inference. Biometrics, 58:21--29

    2002

  13. [13]

    and Melly, B

    Fr\"olich, M. and Melly, B. (2013). Unconditional quantile treatment effects under endogeneity. Journal of Business & Economic Statistics, 31:346--357

    2013

  14. [14]

    Gilbert, P. B. and Hudgens, M. G. (2008). Evaluating candidate principal surrogate endpoints. Biometrics, 64:1146--1154

    2008

  15. [15]

    Hahn, J. (1998). On the role of the propensity score in efficient semiparametric estimation of average treatment effects. Econometrica, 66:315--331

    1998

  16. [16]

    W., and Ridder, G

    Hirano, K., Imbens, G. W., and Ridder, G. (2003). Efficient estimation of average treatment effects using the estimated propensity score. Econometrica, 71:1161--1189

    2003

  17. [17]

    and Ratkovic, M

    Imai, K. and Ratkovic, M. (2014). Covariate balancing propensity score. Journal of the Royal Statistical Society: Series B, 76:243--263

    2014

  18. [18]

    Imbens, G., Kallus, N., Mao, X., and Wang, Y. (2025). Long-term causal inference under persistent confounding via data combination. Journal of the Royal Statistical Society: Series B, 87:362--388

    2025

  19. [19]

    H., Balakrishnan, S., and Wasserman, L

    Kennedy, E. H., Balakrishnan, S., and Wasserman, L. (2023). Semiparametric counterfactual density estimation. Biometrika, 110:875--896

    2023

  20. [20]

    and Bassett, G

    Koenker, R. and Bassett, G. (1978). Regression quantiles. Econometrica, 46:33--50

    1978

  21. [21]

    Lee, D., Yang, S., and Wang, X. (2022). Doubly robust estimators for generalizing treatment effects on survival outcomes from randomized controlled trials to a target population. Journal of Causal Inference, 10:415--440

    2022

  22. [22]

    Newey, W. K. (1994). The asymptotic variance of semiparametric estimators. Econometrica, 62:1349--1382

    1994

  23. [23]

    Newey, W. K. (1997). Convergence rates and asymptotic normality for series estimators. Journal of Econometrics, 79:147--168

    1997

  24. [24]

    Prentice, R. L. (1989). Surrogate endpoints in clinical trials: Definition and operational criteria. Statistics in Medicine, 8:431--440

    1989

  25. [25]

    and Lawless, J

    Qin, J. and Lawless, J. (1994). Empirical likelihood and general estimating equations. The Annals of Statistics, 22:300--325

    1994

  26. [26]

    Rosenbaum, P. R. and Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70:41--55

    1983

  27. [27]

    Rubin, D. B. (1978). Bayesian inference for causal effects: The role of randomization. The Annals of Statistics, 6:34--58

    1978

  28. [28]

    Rubin, D. B. (2005). Causal inference using potential outcomes: Design, modeling, decisions. Journal of the American Statistical Association, 100:322--331

    2005

  29. [29]

    Stone, C. J. (1980). Optimal rates of convergence for nonparametric estimators. The Annals of Statistics, 8:1348--1360

    1980

  30. [30]

    Stone, C. J. (1985). Additive regression and other nonparametric models. The Annals of Statistics, 13:689--705

    1985

  31. [31]

    and Fu, H

    Tao, Y. and Fu, H. (2019). Doubly robust estimation of the weighted average treatment effect for a target population. Statistics in Medicine, 38:315--325

    2019

  32. [32]

    VanderWeele, T. J. (2013). Surrogate measures and consistent surrogates. Biometrics, 69:561--565

    2013

  33. [33]

    and Zhu, Z

    Zhang, Y. and Zhu, Z. (2025). A data fusion method for quantile treatment effects. Statistica Sinica, 35:981--1002

    2025

  34. [34]

    Abadie, A., Angrist, J., and Imbens, G. (2002). Instrumental variables estimates of the effect of subsidized training on the quantiles of trainee earnings. Econometrica, 70:91--117

    2002

  35. [35]

    Belloni, A., Chernozhukov, V., Fern\'andez-Val, I., and Hansen, C. (2017). Program evaluation and causal inference with high-dimensional data. Econometrica, 85:233--298

    2017

  36. [36]

    and De Vito, E

    Caponnetto, A. and De Vito, E. (2007). Optimal rates for the regularized least-squares algorithm. Foundations of Computational Mathematics, 7:331--368

    2007

  37. [37]

    and Hansen, C

    Chernozhukov, V. and Hansen, C. (2005). An IV model of quantile treatment effects. Econometrica, 73:245--261

    2005

  38. [38]

    Chernozhukov, V., Fern\'andez-Val, I., and Galichon, A. (2010). Quantile and probability curves without crossing. Econometrica, 78:1093--1125

    2010

  39. [39]

    Chernozhukov, V., Fern\'andez-Val, I., and Melly, B. (2013). Inference on counterfactual distributions. Econometrica, 81:2205--2268

    2013

  40. [40]

    Chernozhukov, V., Chetverikov, D., and Kato, K. (2014). Gaussian approximation of suprema of empirical processes. The Annals of Statistics, 42:1564--1597

    2014

  41. [41]

    Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W., and Robins, J. (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21:C1--C68

    2018

  42. [42]

    D\'iaz, I. (2017). Efficient estimation of quantiles in missing data models. Journal of Statistical Planning and Inference, 190:39--51

    2017

  43. [43]

    Firpo, S. (2007). Efficient semiparametric estimation of quantile treatment effects. Econometrica, 75:259--276

    2007

  44. [44]

    and Mao, X

    Kallus, N. and Mao, X. (2025). On the role of surrogates in the efficient estimation of treatment effects with limited outcome data. Journal of the Royal Statistical Society: Series B, 87:480--509

    2025

  45. [45]

    Kallus, N., Mao, X., and Uehara, M. (2024). Localized debiased machine learning: Efficient inference on quantile treatment effects and beyond. Journal of Machine Learning Research, 25:1--59

    2024

  46. [46]

    M., Rotnitzky, A., and Zhao, L

    Robins, J. M., Rotnitzky, A., and Zhao, L. P. (1994). Estimation of regression coefficients when some regressors are not always observed. Journal of the American Statistical Association, 89:846--866

    1994

  47. [47]

    Robins, J. M. and Rotnitzky, A. (1995). Semiparametric efficiency in multivariate regression models with missing data. Journal of the American Statistical Association, 90:122--129

    1995

  48. [48]

    Tsiatis, A. A. (2006). Semiparametric Theory and Missing Data. Springer

    2006

  49. [49]

    van der Laan, M. J. and Robins, J. M. (2003). Unified Methods for Censored Longitudinal Data and Causality. Springer

    2003

  50. [50]

    van der Vaart, A. W. (1998). Asymptotic Statistics. Cambridge University Press

    1998

  51. [51]

    van der Vaart, A. W. and Wellner, J. A. (1996). Weak Convergence and Empirical Processes. Springer

    1996

  52. [52]

    M., Katzenstein, D

    Hammer, S. M., Katzenstein, D. A., Hughes, M. D., Gundacker, H., Schooley, R. T., Haubrich, R. H., Henry, W. K., Lederman, M. M., Phair, J. P., Niu, M., Hirsch, M. S., and Merigan, T. C. (1996). A trial comparing nucleoside monotherapy with combination therapy in HIV-infected adults with CD4 cell counts from 200 to 500 per cubic millimeter. New England Jo...

    work page doi:10.1056/nejm199610103351501 1996

  53. [53]

    B., Lu, X., Zhang, M., Davidian, M., and Tsiatis, A

    Juraska, M., Gilbert, P. B., Lu, X., Zhang, M., Davidian, M., and Tsiatis, A. A. (2025). ACTG175: AIDS Clinical Trials Group Study 175. R package documentation in speff2trial: Semiparametric Efficient Estimation for a Two-Sample Treatment Effect, version 1.0.5

    2025