pith. sign in

arxiv: 2510.13397 · v3 · submitted 2025-10-15 · 💻 cs.LG · stat.ML

Assessing the robustness of heterogeneous treatment effects in survival analysis under informative censoring

Yuxin Wang , Dennis Frauen , Jonas Schweisthal , Maresa Schr\"oder , Stefan Feuerriegel This is my paper

Pith reviewed 2026-05-18 07:15 UTC · model grok-4.3

classification 💻 cs.LG stat.ML
keywords informative censoringsurvival analysisconditional average treatment effectpartial identificationmeta-learnerdouble robustnesscausal inferenceheterogeneous treatment effects
0
0 comments X

The pith

Partial identification produces informative bounds on conditional treatment effects in survival analysis despite informative censoring.

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

The paper develops a framework to evaluate how robust estimates of heterogeneous treatment effects are in survival analysis when patients drop out in ways that depend on their outcomes. Instead of relying on the usual assumption that censoring is unrelated to survival given covariates, it uses partial identification to calculate ranges that contain the true conditional average treatment effect. A new meta-learner, SurvB-learner, estimates these ranges and works with any base machine learning model while being doubly robust and nearly as efficient as if the true model were known. Experiments on simulated and real data show that this approach can still find subgroups where treatment works even under biased censoring.

Core claim

The core discovery is an assumption-lean method that derives bounds on the conditional average treatment effect (CATE) for survival outcomes under informative censoring by leveraging partial identification, together with a novel SurvB-learner meta-algorithm that estimates those bounds in a doubly robust and quasi-oracle efficient way for arbitrary machine learning models.

What carries the argument

The SurvB-learner, a model-agnostic meta-learner that combines partial identification bounds with flexible machine learning base learners to estimate intervals for the CATE.

Load-bearing premise

The partial identification bounds derived from the observed data under informative censoring are non-vacuous and correctly contain the true CATE.

What would settle it

A simulation study in which the true CATE is known and the estimated bounds from the SurvB-learner fail to cover it under controlled informative censoring would falsify the method.

Figures

Figures reproduced from arXiv: 2510.13397 by Dennis Frauen, Jonas Schweisthal, Maresa Schr\"oder, Stefan Feuerriegel, Yuxin Wang.

Figure 1
Figure 1. Figure 1: Overview of our framework. (A) Data: A key challenge in survival analysis is censor￾ing, where follow-up information about patient outcomes is incomplete. In practice, this can be due to various reasons such as patients dropping out of clinical studies for side effects [3, 39]. As a result, the exact event time (e.g., of death or disease progression) for censored patients is unknown. (B) Intuition behind o… view at source ↗
Figure 2
Figure 2. Figure 2: Causal graph for survival data under informative censoring. Variables in yellow are ob￾served, while variables in blue are unobserved. Solid edges denote causal relationships, and dashed edges reflect dependence. Note that the absence of arrows encodes causal assumptions, not their presence. 1We deal with the problem of right censoring, which is common in survival analysis settings. We thus assume that eve… view at source ↗
Figure 3
Figure 3. Figure 3: Results for experiments with synthetic data. The experiments serve two purposes: (i) to show that our SurvB-learner obtains values close to the oracle bounds, and (ii) to demonstrate that the plug-in learner is not robust and therefore inferior. Experimental details are in Supple￾ment E. (A) We report the mean ± standard deviation of the RMSE compared to the oracle bound over 5 random runs for different sy… view at source ↗
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p029_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Results for experiments with observational datasets without hidden confounding. The results are shown in the same way as [PITH_FULL_IMAGE:figures/full_fig_p082_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: We selected these subgroups according to the data-driven analysis and the co-alteration [PITH_FULL_IMAGE:figures/full_fig_p084_6.png] view at source ↗
Figure 6
Figure 6. Figure 6: Survival curves of conditional average DFS time. These curves show the expected survival time for SMAD4 (top line), CDK4 (middle line), and MYC (bottom line), for patients with alteration (right) and without alteration (left). The x-axis is time since randomization in months, and the y-axis is the probability of conditional average survival time beyond that time point. The lower and upper bounds (shaded ar… view at source ↗
Figure 7
Figure 7. Figure 7: Single-covariate analysis for OS. The forest plot (left) shows the mean and standard deviation of the estimated lower bounds of the CATE within subgroups defined by baseline char￾acteristics or genetic covariates by bootstrap. For each subgroup on the x-axis, we performed 2000 bootstrap replications. The background scatter points represent the CATE lower bound estimates without bootstrapping. The table (ri… view at source ↗
Figure 8
Figure 8. Figure 8: Data-driven analysis for OS. Using a tailored recursive partition algorithm, we select subgroups with the highest percentage of LBs above zero. Method details are in Supplement F. 87 [PITH_FULL_IMAGE:figures/full_fig_p087_8.png] view at source ↗
read the original abstract

Dropout is common in clinical studies, with up to half of patients leaving early due to side effects or other reasons. When dropout is informative (i.e., dependent on survival time), it introduces censoring bias, because of which treatment effect estimates are also biased. In this paper, we propose an assumption-lean framework to assess the robustness of conditional average treatment effect (CATE) estimates in survival analysis when facing censoring bias. Unlike existing works that rely on strong assumptions, such as non-informative censoring, to obtain point estimation, we use partial identification to derive informative bounds on the CATE. Thereby, our framework helps to identify patient subgroups where treatment is effective despite informative censoring. We further propose a novel model-agnostic meta-learner, called SurvB-learner, to estimate the bounds that can be used in combination with arbitrary machine-learning models, and that has favorable theoretical properties such as double-robustness and quasi-oracle efficiency. We finally demonstrate the effectiveness of our meta-learner across various experiments using both simulated and real-world data.

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

2 major / 2 minor

Summary. The manuscript proposes an assumption-lean framework for assessing the robustness of conditional average treatment effect (CATE) estimates in survival analysis under informative censoring. It replaces point estimation with partial identification to derive informative bounds on the CATE, enabling identification of patient subgroups where treatment remains effective despite censoring bias. The authors introduce the SurvB-learner, a model-agnostic meta-learner that can be combined with arbitrary machine learning models and is claimed to possess double-robustness and quasi-oracle efficiency. The approach is illustrated on simulated data and real-world clinical datasets.

Significance. If the partial identification strategy produces non-vacuous bounds and the theoretical properties of the SurvB-learner are rigorously established, the work would address a practically important problem in clinical survival analysis where dropout rates can reach 50%. The model-agnostic meta-learner design and emphasis on robustness without strong parametric assumptions on the censoring mechanism represent potential strengths for downstream applications in heterogeneous treatment effect estimation.

major comments (2)
  1. [Partial identification framework (around the definition of the SurvB-learner bounds)] The central claim that partial identification yields informative (non-vacuous) bounds on the CATE under informative censoring without the non-informative censoring assumption requires explicit justification. The construction of the identified set for the conditional survival functions (and thus the CATE) must impose some restriction on the dependence between survival time T and censoring time C beyond the observed marginals; otherwise the identified set collapses to the full [0,1] interval for many covariate values, rendering the bounds uninformative for subgroup identification. This issue is load-bearing for the framework's utility and should be addressed with a concrete sensitivity parameter, support restriction, or explicit bounding of the copula in the relevant derivation section.
  2. [Theoretical properties of SurvB-learner] The abstract and introduction assert double-robustness and quasi-oracle efficiency for the SurvB-learner. These properties need to be stated with precise conditions (e.g., which nuisance functions must be consistently estimated, and under what rate conditions the quasi-oracle property holds) and proved in the theoretical analysis section. Without these details it is unclear whether the properties survive the partial identification step or depend on additional modeling choices for the censoring mechanism.
minor comments (2)
  1. [Experiments] Clarify the exact data exclusion rules and preprocessing steps used in the real-world experiments to ensure reproducibility.
  2. [Numerical results] Add a table or figure that directly compares the width of the derived CATE bounds against existing point-estimation methods under varying degrees of informative censoring.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive comments on our manuscript. The feedback highlights important areas for clarification in both the partial identification framework and the theoretical properties of the SurvB-learner. We address each major comment below and will revise the manuscript to incorporate the suggested improvements for greater rigor and transparency.

read point-by-point responses

  1. Referee: [Partial identification framework (around the definition of the SurvB-learner bounds)] The central claim that partial identification yields informative (non-vacuous) bounds on the CATE under informative censoring without the non-informative censoring assumption requires explicit justification. The construction of the identified set for the conditional survival functions (and thus the CATE) must impose some restriction on the dependence between survival time T and censoring time C beyond the observed marginals; otherwise the identified set collapses to the full [0,1] interval for many covariate values, rendering the bounds uninformative for subgroup identification. This issue is load-bearing for the framework's utility and should be addressed with a concrete sensitivity parameter, support restriction, or explicit bounding of the copula in the relevant derivation section.

    Authors: We agree that the current presentation would benefit from more explicit justification to ensure the bounds are non-vacuous. While the framework relies on the observed data distribution to derive the identified set, we acknowledge that without an additional restriction on the dependence between T and C, the bounds could indeed become uninformative in some regions. In the revised manuscript, we will introduce a concrete sensitivity parameter that bounds the possible copula dependence between survival and censoring times. This parameter will be defined and motivated in the derivation section, with explicit conditions under which the resulting CATE bounds remain informative for subgroup identification. We will also include guidance on interpreting and selecting the parameter in practice. revision: yes

  2. Referee: [Theoretical properties of SurvB-learner] The abstract and introduction assert double-robustness and quasi-oracle efficiency for the SurvB-learner. These properties need to be stated with precise conditions (e.g., which nuisance functions must be consistently estimated, and under what rate conditions the quasi-oracle property holds) and proved in the theoretical analysis section. Without these details it is unclear whether the properties survive the partial identification step or depend on additional modeling choices for the censoring mechanism.

    Authors: We agree that the theoretical claims require more precise statements and supporting proofs. The double-robustness of the SurvB-learner is with respect to consistent estimation of the conditional survival functions and the censoring distribution, while the quasi-oracle efficiency holds when the nuisance estimators satisfy appropriate convergence rates (faster than n^{-1/4}). These properties are intended to carry over to the partial identification bounds without requiring parametric assumptions on the censoring mechanism beyond what is identifiable from the data. In the revision, we will expand the theoretical analysis section to state the full set of assumptions, present the relevant theorems with exact conditions, and include the proofs. We will also clarify how the partial identification step interacts with these guarantees. revision: yes

Circularity Check

0 steps flagged

No significant circularity: partial identification framework derives bounds independently of point estimates

full rationale

The derivation relies on partial identification to produce bounds on CATE under informative censoring, replacing point estimation with an identified set that incorporates the observed censoring mechanism without requiring non-informative censoring. This is a standard sensitivity-style relaxation in causal inference and does not reduce to a fitted parameter or self-citation by construction. The SurvB-learner is presented as a model-agnostic meta-learner with double-robustness and quasi-oracle efficiency, properties that follow from standard semiparametric theory rather than re-labeling inputs as outputs. No load-bearing step equates the claimed bounds or theoretical guarantees to the raw data or prior fits; the framework remains self-contained against external benchmarks for partial identification.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the validity of deriving non-trivial bounds via partial identification when censoring is informative and on the double-robustness properties of the newly proposed meta-learner; these are domain assumptions about the data-generating process rather than results derived from first principles or external benchmarks.

axioms (1)
  • domain assumption Informative bounds on the CATE can be derived from the observed data under the censoring mechanism without invoking non-informative censoring.
    This replaces point estimation and is invoked when the framework is described as assumption-lean yet still able to produce informative bounds.
invented entities (1)
  • SurvB-learner no independent evidence
    purpose: Model-agnostic meta-learner for estimating partial identification bounds on CATE
    Newly introduced method whose double-robustness and quasi-oracle efficiency are claimed but rest on the paper's own theoretical analysis.

pith-pipeline@v0.9.0 · 5735 in / 1515 out tokens · 50436 ms · 2026-05-18T07:15:32.298787+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

102 extracted references · 102 canonical work pages · 1 internal anchor

  1. [1]

    Cancer care treatment attrition in adults: Measurement approaches and inequities in patient dropout rates: a rapid review.BMC Cancer, 24(1):1345, 2024

    Jenny Shand, Elizabeth Stovold, Lucy Goulding, and Kate Cheema. Cancer care treatment attrition in adults: Measurement approaches and inequities in patient dropout rates: a rapid review.BMC Cancer, 24(1):1345, 2024

    work page 2024

  2. [2]

    Alekseev, Mustafa Özgüro˘glu, Dingwei Ye, Susan Feyerabend, Andrew Protheroe, Peter De Porre, Thian Kheoh, Youn C

    Karim Fizazi, NamPhuong Tran, Luis Fein, Nobuaki Matsubara, Alfredo Rodriguez- Antolin, Boris Y . Alekseev, Mustafa Özgüro˘glu, Dingwei Ye, Susan Feyerabend, Andrew Protheroe, Peter De Porre, Thian Kheoh, Youn C. Park, Mary B. Todd, Kim N. Chi, and LATITUDE Investigators. Abiraterone plus prednisone in metastatic, castration-sensitive prostate cancer.The ...

    work page 2017

  3. [3]

    Informative censoring in externally controlled clinical trials: a potential source of bias.ESMO Open, 10(1):104094, 2025

    Tulika Rudra Gupta, Daniel Schwartz, Riddhiman Saha, Patrick Wen, Rahman Rahman, and Lorenzo Trippa. Informative censoring in externally controlled clinical trials: a potential source of bias.ESMO Open, 10(1):104094, 2025

    work page 2025

  4. [4]

    Templeton, Eitan Amir, and Ian F

    Arnoud J. Templeton, Eitan Amir, and Ian F. Tannock. Informative censoring — a neglected cause of bias in oncology trials.Nature Reviews Clinical Oncology, 17(6):327–328, 2020

    work page 2020

  5. [5]

    Van Der Laan and James M

    Mark J. Van Der Laan and James M. Robins.Unified Methods for Censored Longitudinal Data and Causality. Springer Series in Statistics. Springer, New York, NY , 2003. ISBN 978-0-387-21700-0

    work page 2003

  6. [6]

    Klein and Melvin L

    John P. Klein and Melvin L. Moeschberger.Survival Analysis: Techniques for Censored and Truncated Data. Statistics for Biology and Health. Springer, New York, 2003

    work page 2003

  7. [7]

    Attrition rates, reasons, and predictive factors in supportive care and palliative oncology clinical trials

    David Hui, Isabella Glitza, Gary Chisholm, Sriram Yennu, and Eduardo Bruera. Attrition rates, reasons, and predictive factors in supportive care and palliative oncology clinical trials. Cancer, 119(5):1098–1105, 2013. 34

    work page 2013

  8. [8]

    Perez-Cruz, Omar Shamieh, Carlos Eduardo Paiva, Jung Hye Kwon, Mary Ann Muckaden, Eduardo Bruera, and David Hui

    Pedro E. Perez-Cruz, Omar Shamieh, Carlos Eduardo Paiva, Jung Hye Kwon, Mary Ann Muckaden, Eduardo Bruera, and David Hui. Factors associated with attrition in a multicen- ter longitudinal observational study of patients with advanced cancer.Journal of Pain and Symptom Management, 55(3):938–945, 2018

    work page 2018

  9. [9]

    E. L. Kaplan and Paul Meier. Nonparametric estimation from incomplete observations. Journal of the American Statistical Association, 53(282):457–481, 1958

    work page 1958

  10. [10]

    D. R. Cox. Regression models and life-tables.Journal of the Royal Statistical Society: Series B (Methodological), 34(2):187–202, 1972

    work page 1972

  11. [11]

    N. E. Breslow. Analysis of survival data under the proportional hazards model.International Statistical Review / Revue Internationale de Statistique, 43(1):45–57, 1975

    work page 1975

  12. [12]

    L. J. Wei. The accelerated failure time model: a useful alternative to the cox regression model in survival analysis.Statistics in Medicine, 11(14-15):1871–1879, 1992

    work page 1992

  13. [13]

    Gao and T

    Zijun Gao and Trevor Hastie. Estimating heterogeneous treatment effects for general re- sponses.arXiv preprint, arXiv:2103.04277, 2022

    work page arXiv 2022

  14. [14]

    Mining heterogeneous causal effects for personalized cancer treatment.Bioinformatics, 33(15):2372–2378, 2017

    Weijia Zhang, Thuc Duy Le, Lin Liu, Zhi-Hua Zhou, and Jiuyong Li. Mining heterogeneous causal effects for personalized cancer treatment.Bioinformatics, 33(15):2372–2378, 2017

    work page 2017

  15. [15]

    Henderson, Thomas A

    Nicholas C. Henderson, Thomas A. Louis, Gary L. Rosner, and Ravi Varadhan. Individ- ualized treatment effects with censored data via fully nonparametric Bayesian accelerated failure time models.Biostatistics, 21(1):50–68, 2020

    work page 2020

  16. [16]

    Non-parametric individual treatment effect estimation for survival data with random forests.Bioinformatics (Oxford, England), 36(2):629–636, 2020

    Sami Tabib and Denis Larocque. Non-parametric individual treatment effect estimation for survival data with random forests.Bioinformatics (Oxford, England), 36(2):629–636, 2020. 35

    work page 2020

  17. [17]

    Yifan Cui, Michael R Kosorok, Erik Sverdrup, Stefan Wager, and Ruoqing Zhu. Estimating heterogeneous treatment effects with right-censored data via causal survival forests.Journal of the Royal Statistical Society Series B: Statistical Methodology, 85(2):179–211, 2023

    work page 2023

  18. [18]

    Oefner, Tim Beißbarth, Rainer Spang, Helena U

    Stefan Schrod, Andreas Schäfer, Stefan Solbrig, Robert Lohmayer, Wolfram Gronwald, Peter J. Oefner, Tim Beißbarth, Rainer Spang, Helena U. Zacharias, and Michael Al- tenbuchinger. BITES: balanced individual treatment effect for survival data.Bioinformatics, 38(Supplement_1):i60–i67, 2022

    work page 2022

  19. [19]

    Katzman, Uri Shaham, Alexander Cloninger, Jonathan Bates, Tingting Jiang, and Yuval Kluger

    Jared L. Katzman, Uri Shaham, Alexander Cloninger, Jonathan Bates, Tingting Jiang, and Yuval Kluger. Deepsurv: personalized treatment recommender system using a cox propor- tional hazards deep neural network.BMC Medical Research Methodology, 18(1):24, 2018

    work page 2018

  20. [20]

    Curth and M

    Alicia Curth and Mihaela van der Schaar. Nonparametric estimation of heterogeneous treat- ment effects: From theory to learning algorithms.arXiv preprint, arXiv:2101.10943, 2021

    work page arXiv 2021

  21. [21]

    Treatment heterogeneity for survival outcomes.arXiv preprint, arXiv:2207.07758, 2022

    Yizhe Xu, Nikolaos Ignatiadis, Erik Sverdrup, Scott Fleming, Stefan Wager, and Nigam Shah. Treatment heterogeneity for survival outcomes.arXiv preprint, arXiv:2207.07758, 2022

    work page arXiv 2022

  22. [22]

    Finkelstein, Roy E

    Shenbo Xu, Raluca Cobzaru, Stan N. Finkelstein, Roy E. Welsch, Kenney Ng, and Zach Shahn. Estimating heterogeneous treatment effects on survival outcomes using counterfac- tual censoring unbiased transformations.arXiv preprint, arXiv:2401.11263, 2024

    work page arXiv 2024

  23. [23]

    Orthogonal survival learners for estimating heterogeneous treatment effects from time-to-event data

    Dennis Frauen, Maresa Schröder, Konstantin Hess, and Stefan Feuerriegel. Orthogonal survival learners for estimating heterogeneous treatment effects from time-to-event data. In NeurIPS, 2025

    work page 2025

  24. [24]

    Schaubel and Guanghui Wei

    Douglas E. Schaubel and Guanghui Wei. Double inverse-weighted estimation of cumulative treatment effects under nonproportional hazards and dependent censoring.Biometrics, 67 (1):29–38, 2011. 36

    work page 2011

  25. [25]

    On the propensity score weighting analy- sis with survival outcome: Estimands, estimation, and inference.Statistics in Medicine, 37 (26):3745–3763, 2018

    Huzhang Mao, Liang Li, Wei Yang, and Yu Shen. On the propensity score weighting analy- sis with survival outcome: Estimands, estimation, and inference.Statistics in Medicine, 37 (26):3745–3763, 2018

    work page 2018

  26. [26]

    Partial causal identification for right censored data with noncom- pliance.Journal of Nonparametric Statistics, 2025

    Yang Bai and Yifan Cui. Partial causal identification for right censored data with noncom- pliance.Journal of Nonparametric Statistics, 2025

    work page 2025

  27. [27]

    Causal survival analysis, estimation of the average treatment effect (ATE): Practical rec- ommendations

    Charlotte V oinot, Clément Berenfeld, Imke Mayer, Bernard Sebastien, and Julie Josse. Causal survival analysis, estimation of the average treatment effect (ATE): Practical rec- ommendations. 2025

    work page 2025

  28. [28]

    Bounding causal effects with an unknown mixture of informative and non- informative censoring.arXiv preprint, arXiv:2411.16902, 2025

    Max Rubinstein, Denis Agniel, Larry Han, Marcela Horvitz-Lennon, and Sharon-Lise Normand. Bounding causal effects with an unknown mixture of informative and non- informative censoring.arXiv preprint, arXiv:2411.16902, 2025

    work page arXiv 2025

  29. [29]

    van der Laan

    Daniel Rubin and Mark J. van der Laan. A doubly robust censoring unbiased transformation. The International Journal of Biostatistics, 3(1):Article 4, 2007

    work page 2007

  30. [30]

    One-step Targeted Maximum Likelihood for Time-to-event Outcomes

    Weixin Cai and Mark J. van der Laan. One-step targeted maximum likelihood for time-to- event outcomes.arXiv preprint, arXiv:1802.09479, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  31. [31]

    Addressing extreme propensity scores in estimating counterfactual survival functions via the overlap weights.American Journal of Epidemiology, 191(6):1140–1151, 2022

    Chao Cheng, Fan Li, Laine E Thomas, and Fan (Frank) Li. Addressing extreme propensity scores in estimating counterfactual survival functions via the overlap weights.American Journal of Epidemiology, 191(6):1140–1151, 2022

    work page 2022

  32. [32]

    Gilbert, and Marco Carone

    Ted Westling, Alex Luedtke, Peter B. Gilbert, and Marco Carone. Inference for treatment- specific survival curves using machine learning.Journal of the American Statistical Associ- ation, 119(546):1541–1553, 2024

    work page 2024

  33. [33]

    Estimating heterogeneous survival treatment effect in observational data using machine learning.Statistics in Medicine, 40(21):4691–4713, 2021

    Liangyuan Hu, Jiayi Ji, and Fan Li. Estimating heterogeneous survival treatment effect in observational data using machine learning.Statistics in Medicine, 40(21):4691–4713, 2021. 37

    work page 2021

  34. [34]

    Basu and John P

    Asit P. Basu and John P. Klein. Some recent results in competing risks theory. InInstitute of Mathematical Statistics Lecture Notes - Monograph Series, pages 216–229. Institute of Mathematical Statistics, Hayward, CA, 1982. ISBN 978-0-940600-02-7

    work page 1982

  35. [35]

    Slud and Lawrence V

    Eric V . Slud and Lawrence V . Rubinstein. Dependent competing risks and summary survival curves.Biometrika, 70(3):643–649, 1983

    work page 1983

  36. [36]

    Ming Zheng and John P. Klein. Estimates of marginal survival for dependent competing risks based on an assumed copula.Biometrika, 82(1):127–138, 1995

    work page 1995

  37. [37]

    Wen-Zhao Zhong, Qun Wang, Wei-Min Mao, Song-Tao Xu, Lin Wu, Yi Shen, Yong-Yu Liu, Chun Chen, Ying Cheng, Lin Xu, Jun Wang, Ke Fei, Xiao-Fei Li, Jian Li, Cheng Huang, Zhi-Dong Liu, Shun Xu, Ke-Neng Chen, Shi-Dong Xu, Lun-Xu Liu, Ping Yu, Bu- Hai Wang, Hai-Tao Ma, Hong-Hong Yan, Xue-Ning Yang, Qing Zhou, Yi-Long Wu, Qun Wang, Wei-Min Mao, Lin Wu, Yi Shen, Y...

    work page 2018

  38. [38]

    Genomic signatures define three subtypes of EGFR-mutant stage II–III non-small-cell lung cancer with distinct adjuvant therapy out- comes.Nature Communications, 12(1):6450, 2021

    Si-Yang Liu, Hua Bao, Qun Wang, Wei-Min Mao, Yedan Chen, Xiaoling Tong, Song-Tao Xu, Lin Wu, Yu-Cheng Wei, Yong-Yu Liu, Chun Chen, Ying Cheng, Rong Yin, Fan Yang, Sheng-Xiang Ren, Xiao-Fei Li, Jian Li, Cheng Huang, Zhi-Dong Liu, Shun Xu, Ke-Neng Chen, Shi-Dong Xu, Lun-Xu Liu, Ping Yu, Bu-Hai Wang, Hai-Tao Ma, Hong-Hong Yan, Song Dong, Xu-Chao Zhang, Jian ...

    work page 2021

  39. [39]

    Deep learning for survival analysis: a review.Artificial Intelligence Review, 57(3):65, 2024

    Simon Wiegrebe, Philipp Kopper, Raphael Sonabend, Bernd Bischl, and Andreas Bender. Deep learning for survival analysis: a review.Artificial Intelligence Review, 57(3):65, 2024

    work page 2024

  40. [40]

    Survite: Learning heterogeneous treatment effects from time-to-event data

    Alicia Curth, Changhee Lee, and Mihaela van der Schaar. Survite: Learning heterogeneous treatment effects from time-to-event data. InNeurIPS, 2021

    work page 2021

  41. [41]

    Flaig, Fabio Franke, Oscar B

    Fred Saad, Eleni Efstathiou, Gerhardt Attard, Thomas W. Flaig, Fabio Franke, Oscar B. Goodman, Stéphane Oudard, Thomas Steuber, Hiroyoshi Suzuki, Daphne Wu, Kesav Yeruva, Peter De Porre, Sabine Brookman-May, Susan Li, Jinhui Li, Shibu Thomas, Kather- ine B. Bevans, Suneel D. Mundle, Sharon A. McCarthy, Dana E. Rathkopf, and ACIS Investigators. Apalutamide...

    work page 2021

  42. [42]

    Donald B. Rubin. Estimating causal effects of treatments in randomized and nonrandomized studies.Journal of Educational Psychology, 66(5):688, 1974

    work page 1974

  43. [43]

    Alden, Biagio Ricciuti, Joao V

    Federica Pecci, Rohit Thummalapalli, Stephanie L. Alden, Biagio Ricciuti, Joao V . Alessi, Arielle Elkrief, Hira Rizvi, Xinan Wang, Mark Jeng, Jacklynn V . Egger, Victor R. Vaz, Adri- ana Barrichello, Giuseppe Lamberti, Alessandro Di Federico, Valentina Santo, Guilherme Rossato de Almeida, Malini Gandhi, Phoebe Clark, Mizuki Nishino, Bruce E. Johnson, Mat...

    work page 2025

  44. [44]

    Detterbeck, Daniel J

    Frank C. Detterbeck, Daniel J. Boffa, Anthony W. Kim, and Lynn T. Tanoue. The eighth edition lung cancer stage classification.Chest, 151(1):193–203, 2017. 39

    work page 2017

  45. [45]

    Reevaluating disease-free survival as an endpoint vs overall survival in stage III adjuvant colon cancer trials.JNCI: Journal of the National Cancer Institute, 114(1):60–67, 2022

    Jun Yin, Mohamed E Salem, Jesse G Dixon, Zhaohui Jin, Romain Cohen, Aimery DeGra- mont, Eric Van Cutsem, Julien Taieb, Steven R Alberts, Norman Wolmark, Hans-Joachim Schmoll, Leonard B Saltz, Thomas J George, Richard R M Goldberg, Rachel Kerr, Sara Lonardi, Takayuki Yoshino, Greg Yothers, Axel Grothey, Thierry Andre, and Qian Shi. Reevaluating disease-fre...

    work page 2022

  46. [46]

    Yusuke Inoue, Yoshihiro Kitahara, Masato Karayama, Kazuhiro Asada, Koji Nishi- moto, Shun Matsuura, Dai Hashimoto, Masato Fujii, Takashi Matsui, Nao Inami, Mikio Toyoshima, Hiroyuki Matsuda, Masaki Ikeda, Mitsuru Niwa, Yusuke Kaida, Masaki Sato, Yasuhiro Ito, Hideki Yasui, Yuzo Suzuki, Hironao Hozumi, Kazuki Furuhashi, Noriyuki Enomoto, Tomoyuki Fujisawa,...

    work page 2025

  47. [47]

    Treatment- free survival after discontinuation of immune checkpoint inhibitors in mnsclc: a systematic review and meta-analysis.Frontiers in Immunology, 14, 2023

    Yue Hu, Shan Liu, Lixing Wang, Yu Liu, Duohan Zhang, and Yinlong Zhao. Treatment- free survival after discontinuation of immune checkpoint inhibitors in mnsclc: a systematic review and meta-analysis.Frontiers in Immunology, 14, 2023

    work page 2023

  48. [48]

    Guido W. Imbens. Nonparametric estimation of average treatment effects under exogeneity: A review.The Review of Economics and Statistics, 86(1):4–29, 2004

    work page 2004

  49. [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. InICML, 2017

    work page 2017

  50. [50]

    Conformalized survival analysis.Journal of the Royal Statistical Society Series B: Statistical Methodology, 85(1):24–45, 2023

    Emmanuel Candès, Lihua Lei, and Zhimei Ren. Conformalized survival analysis.Journal of the Royal Statistical Society Series B: Statistical Methodology, 85(1):24–45, 2023. 40

    work page 2023

  51. [51]

    Hernán, Sonia Hernández-Díaz, and James M

    Miguel A. Hernán, Sonia Hernández-Díaz, and James M. Robins. A structural approach to selection bias.Epidemiology, 15(5):615–625, 2004

    work page 2004

  52. [52]

    Charles F. Manski. Nonparametric bounds on treatment effects.The American Economic Review, 80(2):319–323, 1990

    work page 1990

  53. [53]

    Edward H. Kennedy. Semiparametric doubly robust targeted double machine learning: a review.arXiv preprint, arXiv:2203.06469, 2023

    work page arXiv 2023

  54. [54]

    Künzel, Jasjeet S

    Sören R. Künzel, Jasjeet S. Sekhon, Peter J. Bickel, and Bin Yu. Meta-learners for estimat- ing heterogeneous treatment effects using machine learning.Proceedings of the National Academy of Sciences, 116(10):4156–4165, 2019

    work page 2019

  55. [55]

    Quasi-oracle estimation of heterogeneous treatment effects

    Xinkun Nie and Stefan Wager. Quasi-oracle estimation of heterogeneous treatment effects. arXiv preprint, arXiv:1712.04912, 2020

    work page arXiv 2020

  56. [56]

    Edward H. Kennedy. Towards optimal doubly robust estimation of heterogeneous causal effects.arXiv preprint, arXiv:2004.14497, 2023

    work page arXiv 2004

  57. [57]

    Charles J. Stone. Optimal rates of convergence for nonparametric estimators.The Annals of Statistics, 8(6):1348–1360, 1980

    work page 1980

  58. [58]

    B-learner: Quasi-oracle bounds on heterogeneous causal effects under hidden con- founding

    Miruna Oprescu, Jacob Dorn, Marah Ghoummaid, Andrew Jesson, Nathan Kallus, and Uri Shalit. B-learner: Quasi-oracle bounds on heterogeneous causal effects under hidden con- founding. InICML, 2023

    work page 2023

  59. [59]

    Random forests.Machine Learning, 45(1):5–32, 2001

    Leo Breiman. Random forests.Machine Learning, 45(1):5–32, 2001

    work page 2001

  60. [60]

    Dahabreh, Sarah E

    Issa J. Dahabreh, Sarah E. Robertson, Eric J. Tchetgen, Elizabeth A. Stuart, and Miguel A. Hernán. Generalizing causal inferences from individuals in randomized trials to all trial- eligible individuals.Biometrics, 75(2):685–694, 2019. 41

    work page 2019

  61. [61]

    Heagerty, and Issa J

    Guanbo Wang, Patrick J. Heagerty, and Issa J. Dahabreh. Using effect scores to characterize heterogeneity of treatment effects.JAMA, 331(14):1225–1226, 2024

    work page 2024

  62. [62]

    Janick Weberpals, Stefan Feuerriegel, Mihaela van der Schaar, and Kenneth L. Kehl. Op- portunities for causal machine learning in precision oncology.NEJM AI, 2(8):AIp2500277, 2025

    work page 2025

  63. [63]

    Kohane, and Mihaela van der Schaar

    Stefan Feuerriegel, Dennis Frauen, Valentyn Melnychuk, Jonas Schweisthal, Konstantin Hess, Alicia Curth, Stefan Bauer, Niki Kilbertus, Isaac S. Kohane, and Mihaela van der Schaar. Causal machine learning for predicting treatment outcomes.Nature Medicine, 30 (4):958–968, 2024

    work page 2024

  64. [64]

    A path-sampling method to partially identify causal effects in instrumen- tal variable models.arXiv preprint, arXiv:1910.09502, 2020

    Florian Gunsilius. A path-sampling method to partially identify causal effects in instrumen- tal variable models.arXiv preprint, arXiv:1910.09502, 2020

    work page arXiv 1910

  65. [65]

    A class of algorithms for general instru- mental variable models

    Niki Kilbertus, Matt J Kusner, and Ricardo Silva. A class of algorithms for general instru- mental variable models. InAdvances in Neural Information Processing Systems, volume 33, pages 20108–20119, 2020

    work page 2020

  66. [66]

    Meta- learners for partially-identified treatment effects across multiple environments

    Jonas Schweisthal, Dennis Frauen, Mihaela Van Der Schaar, and Stefan Feuerriegel. Meta- learners for partially-identified treatment effects across multiple environments. InICML, 2024

    work page 2024

  67. [67]

    Learning representations of instruments for partial identification of treat- ment effects

    Jonas Schweisthal, Dennis Frauen, Maresa Schröder, Konstantin Hess, Niki Kilbertus, and Stefan Feuerriegel. Learning representations of instruments for partial identification of treat- ment effects. InICML, 2025

    work page 2025

  68. [68]

    An automated approach to causal inference in discrete settings.Journal of the American Statistical Association, 119(547):1778–1793, 2024

    Guilherme Duarte, Noam Finkelstein, Dean Knox, Jonathan Mummolo, and Ilya Shpitser. An automated approach to causal inference in discrete settings.Journal of the American Statistical Association, 119(547):1778–1793, 2024. 42

    work page 2024

  69. [69]

    The causal-neural con- nection: Expressiveness, learnability, and inference

    Kevin Xia, Kai-Zhan Lee, Yoshua Bengio, and Elias Bareinboim. The causal-neural con- nection: Expressiveness, learnability, and inference. InNeurIPS, 2021

    work page 2021

  70. [70]

    Neural causal models for counterfactual identification and estimation

    Kevin Xia, Yushu Pan, and Elias Bareinboim. Neural causal models for counterfactual identification and estimation. InICLR, 2023

    work page 2023

  71. [71]

    Stochastic causal programming for bounding treatment effects

    Kirtan Padh, Jakob Zeitler, David Watson, Matt Kusner, Ricardo Silva, and Niki Kilbertus. Stochastic causal programming for bounding treatment effects. InCLeaR, 2023

    work page 2023

  72. [72]

    P. R. Rosenbaum and D. B. Rubin. Assessing sensitivity to an unobserved binary covariate in an observational study with binary outcome.Journal of the Royal Statistical Society. Series B (Methodological), 45(2):212–218, 1983

    work page 1983

  73. [73]

    Rosenbaum

    Paul R. Rosenbaum. Sensitivity analysis for certain permutation inferences in matched observational studies.Biometrika, 74(1):13–26, 1987

    work page 1987

  74. [74]

    Siyu Heng and Dylan S. Small. Sharpening the rosenbaum sensitivity bounds to address concerns about interactions between observed and unobserved covariates.Statistica Sinica, 31:2331–2353, 2021

    work page 2021

  75. [75]

    A distributional approach for causal inference using propensity scores.Jour- nal of the American Statistical Association, 101(476):1619–1637, 2006

    Zhiqiang Tan. A distributional approach for causal inference using propensity scores.Jour- nal of the American Statistical Association, 101(476):1619–1637, 2006

    work page 2006

  76. [76]

    Interval estimation of individual-level causal effects under unobserved confounding

    Nathan Kallus, Xiaojie Mao, and Angela Zhou. Interval estimation of individual-level causal effects under unobserved confounding. InAISTATS, 2019

    work page 2019

  77. [77]

    Small, and Bhaswar B

    Qingyuan Zhao, Dylan S. Small, and Bhaswar B. Bhattacharya. Sensitivity analysis for inverse probability weighting estimators via the percentile bootstrap.Journal of the Royal Statistical Society Series B: Statistical Methodology, 81(4):735–761, 2019

    work page 2019

  78. [78]

    Quantifying ignorance in individual-level causal-effect estimates under hidden confounding

    Andrew Jesson, Sören Mindermann, Yarin Gal, and Uri Shalit. Quantifying ignorance in individual-level causal-effect estimates under hidden confounding. InICML, 2021. 43

    work page 2021

  79. [79]

    Sharp sensitivity analysis for inverse propensity weighting via quantile balancing.Journal of the American Statistical Association, 118(544):2645–2657, 2023

    Jacob Dorn and Kevin Guo. Sharp sensitivity analysis for inverse propensity weighting via quantile balancing.Journal of the American Statistical Association, 118(544):2645–2657, 2023

    work page 2023

  80. [80]

    Sharp bounds for generalized causal sensitivity analysis

    Dennis Frauen, Valentyn Melnychuk, and Stefan Feuerriegel. Sharp bounds for generalized causal sensitivity analysis. InNeurIPS, 2023

    work page 2023

Showing first 80 references.