pith. sign in

arxiv: 2503.14459 · v2 · submitted 2025-03-18 · 📊 stat.ML · cs.LG· stat.ME

Doubly robust identification of treatment effects from multiple environments

Piersilvio De Bartolomeis , Julia Kostin , Javier Abad , Yixin Wang , Fanny Yang This is my paper

Pith reviewed 2026-05-22 23:35 UTC · model grok-4.3

classification 📊 stat.ML cs.LGstat.ME
keywords causal inferencetreatment effect estimationmultiple environmentsdoubly robust identificationobservational datainvariance assumptionheterogeneity
0
0 comments X

The pith

RAMEN identifies treatment effects from multiple data sources without the causal graph by using double robustness from invariance.

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

The paper introduces RAMEN, an algorithm for estimating treatment effects from observational data collected across multiple environments. It leverages heterogeneity among these sources to produce unbiased estimates without requiring knowledge or learning of the underlying causal graph. The central property is double robustness: the average treatment effect is identifiable if the causal parents of either the treatment or the outcome are observed, provided that node satisfies an invariance assumption across environments. This approach targets settings in medicine and social sciences where randomized trials are impractical and full causal graphs are unavailable.

Core claim

RAMEN achieves doubly robust identification of treatment effects from multiple environments: the treatment effect is identifiable whenever the causal parents of the treatment or those of the outcome are observed, and the node whose parents are observed satisfies an invariance assumption.

What carries the argument

Doubly robust identification that exploits observed causal parents of the treatment or outcome satisfying an invariance assumption across heterogeneous data sources.

If this is right

  • The treatment effect remains identifiable even if only the parents of the treatment satisfy the conditions.
  • The treatment effect remains identifiable even if only the parents of the outcome satisfy the conditions.
  • No knowledge or recovery of the full causal graph is required for valid estimation.
  • The method applies directly to observational datasets in medicine and social sciences where post-treatment or unobserved variables may be present.

Where Pith is reading between the lines

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

  • The same invariance logic might extend to partial identification when some but not all relevant parents are observed.
  • Adding more environments could tighten bounds or relax the required heterogeneity level.
  • The approach could be tested by constructing synthetic environments with controlled invariance violations.

Load-bearing premise

The multiple data sources must exhibit sufficient heterogeneity and the invariance assumption must hold for the observed parents node.

What would settle it

A collection of environments where the invariance assumption holds for the relevant parents node yet RAMEN's estimate differs from the true effect recovered by a randomized experiment on the same variables.

Figures

Figures reproduced from arXiv: 2503.14459 by Fanny Yang, Javier Abad, Julia Kostin, Piersilvio De Bartolomeis, Yixin Wang.

Figure 1
Figure 1. Figure 1: Two causal graphs illustrating when the set of all covariates is or is not a valid adjustment set: [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (Row 1) For all the plots: n = 2500, d =5, |E| =5. We plot the mean absolute error averaged across environments when: (a) both invariances are preserved; (b) the invariance w.r.t Y is preserved; (c) the invariance w.r.t T is preserved. We report mean and standard error over 20 runs. (Row 2) Graphical models that capture our data generating process: (a) U does not break any invariance; (b) U breaks the inva… view at source ↗
Figure 4
Figure 4. Figure 4: Mean absolute error aver 537 independent noise, or wher iidd iA 537 treatment feature is either a independent noiseor where 534 ✓null performs competitively limited impact on the outco 533 ✓null performs competitively [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: Mean absolute error aver he post 537 dataset when d independent noise, or where neither T nor Y remains in 537 aged across env dataset when d treatment feature is either a descendant of the outcome, independent noiseor where neither T nor Y remains in estingly 534 ✓null performs competitively since the confounders have a limited impact on the outcome and treatment assignment 533 ✓null performs competiti… view at source ↗
Figure 5
Figure 5. Figure 5: Although neither full set of parents is observed, one can still find a valid adjustment set {X1, X2} (de￾picted in green). First, we observe here that Assumption 3.2 is not a minimal “observabil￾ity” condition on the parents of Y and T: in some cases, it might still be possible to find a valid adjustment set via the observed parents of either T or Y (or both), although no full set of parents was observed (… view at source ↗
Figure 4
Figure 4. Figure 4: Mean absolute error aver the outcome and treatment assignment Additional experiments where the post 537 independent noise, or where neither T nor Y 536 treatment feature is either a descendant of th iddihihT Y ironments, and adjusting for all features ltifIttil 533 534 ✓null performs competitively since the confoun 533 (✓all) generally results in poor performance. In ˆ ✓ ll performs competitively since t… view at source ↗
Figure 4
Figure 4. Figure 4: Mean absolute error aver 537 independent noise, or wher iidd iA 537 treatment feature is either a independent noiseor where 534 ✓null performs competitively limited impact on the outco 533 ✓null performs competitively [PITH_FULL_IMAGE:figures/full_fig_p028_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: Mean absolute error aver 537 independent noise, or wher iidd iA 537 treatment feature is either a independent noiseor where 534 ✓null performs competitively limited impact on the outco 533 ✓null performs competitively invariance only(c) Tinvariance [PITH_FULL_IMAGE:figures/full_fig_p029_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: Mean ab ome and treatment assignment l experiments where the post ome and treatment assignment s, and adjusting for all features her MAE when Y is not in s, and adjusting for all features pp ment; we report mean and standard er is preserved. We plot the mean absolute [PITH_FULL_IMAGE:figures/full_fig_p030_4.png] view at source ↗
read the original abstract

Practical and ethical constraints often require the use of observational data for causal inference, particularly in medicine and social sciences. Yet, observational datasets are prone to confounding, potentially compromising the validity of causal conclusions. While it is possible to correct for biases if the underlying causal graph is known, this is rarely a feasible ask in practical scenarios. A common strategy is to adjust for all available covariates, yet this approach can yield biased treatment effect estimates, especially when post-treatment or unobserved variables are present. We propose RAMEN, an algorithm that produces unbiased treatment effect estimates by leveraging the heterogeneity of multiple data sources without the need to know or learn the underlying causal graph. Notably, RAMEN achieves doubly robust identification: it can identify the treatment effect whenever the causal parents of the treatment or those of the outcome are observed, and the node whose parents are observed satisfies an invariance assumption. Empirical evaluations on synthetic and real-world datasets show that our approach outperforms existing methods.

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 / 1 minor

Summary. The paper introduces RAMEN, an algorithm that produces unbiased treatment effect estimates from multiple observational data sources by exploiting heterogeneity across environments, without requiring knowledge or learning of the underlying causal graph. It claims a doubly robust identification result: the average treatment effect is identified whenever the causal parents of the treatment or of the outcome are observed and the node with observed parents satisfies an invariance assumption. The approach is evaluated empirically on synthetic and real-world datasets, where it outperforms existing methods.

Significance. If the doubly robust identification result holds under the stated conditions on invariance and heterogeneity, the work would offer a practically useful advance in causal inference. It enables graph-free estimation from multi-environment observational data, which is common in medicine and social sciences, while providing robustness to misspecification of either the treatment or outcome mechanism. The empirical outperformance on both synthetic and real data strengthens the case for its utility.

minor comments (1)
  1. [Abstract] Abstract, final sentence of contribution description: the invariance assumption and the precise heterogeneity conditions across environments are referenced but not defined; a one-sentence clarification of these would improve readability without altering the central claim.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our work on RAMEN and for recommending minor revision. The recognition of the practical utility of the doubly robust identification result from multi-environment data without requiring the causal graph is appreciated. No major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract presents RAMEN as achieving doubly robust identification of treatment effects from multiple environments under an invariance assumption on observed causal parents, without any displayed equations, fitted parameters, or derivation steps that reduce the claimed result to its own inputs by construction. No self-definitional loops, fitted-input predictions, or load-bearing self-citations are visible in the supplied text. The central claim is framed as relying on external heterogeneity across data sources and the stated invariance condition, rendering the derivation self-contained on the given information.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the invariance assumption and data heterogeneity across environments, which are stated as domain assumptions rather than derived; no free parameters or invented entities beyond the algorithm itself are mentioned.

axioms (2)
  • domain assumption Invariance assumption holds for the node whose parents are observed
    Required for the doubly robust identification property as stated in the abstract.
  • domain assumption Multiple data sources exhibit heterogeneity sufficient for identification
    Leveraged to achieve identification without the causal graph.
invented entities (1)
  • RAMEN algorithm no independent evidence
    purpose: Produces unbiased treatment effect estimates from multiple environments
    Newly proposed method whose properties are claimed in the abstract.

pith-pipeline@v0.9.0 · 5704 in / 1183 out tokens · 40799 ms · 2026-05-22T23:35:35.377028+00:00 · methodology

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

100 extracted references · 100 canonical work pages · 1 internal anchor

  1. [1]

    Explaining causal findings without bias: detecting and assessing direct effects

    Avidit Acharya, Matthew Blackwell, and Maya Sen. Explaining causal findings without bias: detecting and assessing direct effects. American Political Science Review, 110(3):512–529, 2016

    work page 2016

  2. [2]

    The costs of low birth weight

    Douglas Almond, Kenneth Chay, and David Lee. The costs of low birth weight. The Quarterly Journal of Economics, 120(3):1031–1083, 2005

    work page 2005

  3. [3]

    Invariant Risk Minimization

    Martin Arjovsky, L´ eon Bottou, Ishaan Gulrajani, and David Lopez-Paz. Invariant risk minimization. arXiv preprint arXiv:1907.02893 , 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  4. [4]

    Doubly robust identification for causal panel data models

    Dmitry Arkhangelsky and Guido Imbens. Doubly robust identification for causal panel data models. The Econometrics Journal , 25(3):649–674, 2022

    work page 2022

  5. [5]

    An introduction to propensity score methods for reducing the effects of confounding in observational studies

    Peter Austin. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behavioral Research, 46(3):399–424, 2011

    work page 2011

  6. [6]

    Half-trek criterion for identifiability of latent variable models

    Rina Foygel Barber, Mathias Drton, Nils Sturma, and Luca Weihs. Half-trek criterion for identifiability of latent variable models. The Annals of Statistics , 50(6):3174–3196, 2022

    work page 2022

  7. [7]

    The moderator–mediator variable distinction in social psychologi- cal research: Conceptual, strategic, and statistical considerations

    Reuben Baron and David Kenny. The moderator–mediator variable distinction in social psychologi- cal research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6):1173, 1986

    work page 1986

  8. [8]

    Efficient semiparametric estimation of multi-valued treatment effects under ignora- bility

    Matias Cattaneo. Efficient semiparametric estimation of multi-valued treatment effects under ignora- bility. Journal of Econometrics , 155(2):138–154, 2010

    work page 2010

  9. [9]

    Causal query in observational data with hidden variables

    Debo Cheng, Jiuyong Li, Lin Liu, Jixue Liu, Kui Yu, and Thuc Duy Le. Causal query in observational data with hidden variables. European Conference on Artificial Intelligence, 2020

    work page 2020

  10. [10]

    Toward unique and unbiased causal effect estimation from data with hidden variables

    Debo Cheng, Jiuyong Li, Lin Liu, Kui Yu, Thuc Duy Le, and Jixue Liu. Toward unique and unbiased causal effect estimation from data with hidden variables. IEEE Transactions on Neural Networks and Learning Systems, 34(9):6108–6120, 2022

    work page 2022

  11. [11]

    Local search for efficient causal effect estimation

    Debo Cheng, Jiuyong Li, Lin Liu, Jiji Zhang, Jixue Liu, and Thuc Duy Le. Local search for efficient causal effect estimation. IEEE Transactions on Knowledge and Data Engineering , 2022

    work page 2022

  12. [12]

    Data-driven causal effect estimation based on graphical causal modelling: A survey

    Debo Cheng, Jiuyong Li, Lin Liu, Jixue Liu, and Thuc Duy Le. Data-driven causal effect estimation based on graphical causal modelling: A survey. ACM Computing Surveys , 56(5):1–37, 2024

    work page 2024

  13. [13]

    Double/debiased machine learning for treatment and structural parameters

    Victor Chernozhukov, Denis Chetverikov, Mert Demirer, Esther Duflo, Christian Hansen, Whitney Newey, and James Robins. Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal , 21(1):C1–C68, 2018. 12

    work page 2018

  14. [14]

    Hidden yet quantifi- able: A lower bound for confounding strength using randomized trials

    Piersilvio De Bartolomeis, Javier Abad, Konstantin Donhauser, and Fanny Yang. Hidden yet quantifi- able: A lower bound for confounding strength using randomized trials. International Conference on Artificial Intelligence and Statistics , 2024

    work page 2024

  15. [15]

    Detecting critical treatment effect bias in small subgroups

    Piersilvio De Bartolomeis, Javier Abad, Konstantin Donhauser, and Fanny Yang. Detecting critical treatment effect bias in small subgroups. Uncertainty in Artificial Intelligence , 2024

    work page 2024

  16. [16]

    Covariate selection for the nonpara- metric estimation of an average treatment effect

    Xavier De Luna, Ingeborg Waernbaum, and Thomas Richardson. Covariate selection for the nonpara- metric estimation of an average treatment effect. Biometrika, 98(4):861–875, 2011

    work page 2011

  17. [17]

    Benchmarking observational studies with experimental data under right-censoring

    Ilker Demirel, Edward De Brouwer, Zeshan Hussain, Michael Oberst, Anthony Philippakis, and David Sontag. Benchmarking observational studies with experimental data under right-censoring. arXiv preprint arXiv:2402.15137, 2024

    work page arXiv 2024

  18. [18]

    Npci: Non-parametrics for causal inference

    Vincent Dorie. Npci: Non-parametrics for causal inference. 2016. URL https://github.com/vdorie/ npci

    work page 2016

  19. [19]

    Global identifiability of linear structural equation models

    Mathias Drton, Rina Foygel, and Seth Sullivant. Global identifiability of linear structural equation models. The Annals of Statistics , 39(2):865–886, 2011

    work page 2011

  20. [20]

    Data-driven covariate selection for nonparametric esti- mation of causal effects

    Doris Entner, Patrik Hoyer, and Peter Spirtes. Data-driven covariate selection for nonparametric esti- mation of causal effects. International Conference on Artificial Intelligence and Statistics , 2013

    work page 2013

  21. [21]

    IDA with background knowledge.Uncertainty in Artificial Intelligence, 2020

    Zhuangyan Fang and Yangbo He. IDA with background knowledge.Uncertainty in Artificial Intelligence, 2020

    work page 2020

  22. [22]

    Half-trek criterion for generic identifiability of linear structural equation models

    Rina Foygel, Jan Draisma, and Mathias Drton. Half-trek criterion for generic identifiability of linear structural equation models. The Annals of Statistics , pages 1682–1713, 2012

    work page 2012

  23. [23]

    Learning causal structures using regression invariance

    AmirEmad Ghassami, Saber Salehkaleybar, Negar Kiyavash, and Kun Zhang. Learning causal structures using regression invariance. Advances in Neural Information Processing Systems , 2017

    work page 2017

  24. [24]

    A kernel two-sample test

    Arthur Gretton, Karsten Borgwardt, Malte Rasch, Bernhard Sch¨ olkopf, and Alexander Smola. A kernel two-sample test. The Journal of Machine Learning Research , 13(1):723–773, 2012

    work page 2012

  25. [25]

    arXiv preprint arXiv:2405.04715

    Yihong Gu, Cong Fang, Peter B¨ uhlmann, and Jianqing Fan. Causality pursuit from heterogeneous environments via neural adversarial invariance learning. arXiv preprint arXiv:2405.04715 , 2024

    work page arXiv 2024

  26. [26]

    Differentiable causal backdoor dis- covery

    Limor Gultchin, Matt Kusner, Varun Kanade, and Ricardo Silva. Differentiable causal backdoor dis- covery. International Conference on Artificial Intelligence and Statistics , 2020

    work page 2020

  27. [27]

    Confounder selection: Objectives and ap- proaches

    Richard Guo, Anton Rask Lundborg, and Qingyuan Zhao. Confounder selection: Objectives and ap- proaches. arXiv preprint arXiv:2208.13871 , 2022

    work page arXiv 2022

  28. [28]

    Variable elimination, graph reduction and the efficient g-formula

    Richard Guo, Emilija Perkovi´ c, and Andrea Rotnitzky. Variable elimination, graph reduction and the efficient g-formula. Biometrika, 110(3):739–761, 2023

    work page 2023

  29. [29]

    Confidence intervals for causal effects with invalid instruments by using two-stage hard thresholding with voting

    Zijian Guo, Hyunseung Kang, Tony Cai, and Dylan Small. Confidence intervals for causal effects with invalid instruments by using two-stage hard thresholding with voting. Journal of the Royal Statistical Society Series B: Statistical Methodology , 80(4):793–815, 2018

    work page 2018

  30. [30]

    Functional restriction and efficiency in causal inference

    Jinyong Hahn. Functional restriction and efficiency in causal inference. The Review of Economics and Statistics, 86(1):73–76, 2004

    work page 2004

  31. [31]

    Valid causal inference with (some) invalid instruments

    Jason Hartford, Victor Veitch, Dhanya Sridhar, and Kevin Leyton-Brown. Valid causal inference with (some) invalid instruments. International Conference on Machine Learning , 2021. 13

    work page 2021

  32. [32]

    Robust inference in summary data mendelian randomization via the zero modal pleiotropy assumption

    Fernando Pires Hartwig, George Davey Smith, and Jack Bowden. Robust inference in summary data mendelian randomization via the zero modal pleiotropy assumption. International Journal of Epidemi- ology, 46(6):1985–1998, 2017

    work page 1985

  33. [33]

    Invariant causal prediction for nonlinear models

    Christina Heinze-Deml, Jonas Peters, and Nicolai Meinshausen. Invariant causal prediction for nonlinear models. Journal of Causal Inference , 6(2):20170016, 2018

    work page 2018

  34. [34]

    Graphical criteria for efficient total effect estimation via adjustment in causal linear models

    Leonard Henckel, Emilija Perkovi´ c, and Marloes Maathuis. Graphical criteria for efficient total effect estimation via adjustment in causal linear models. Journal of the Royal Statistical Society Series B: Statistical Methodology, 84(2):579–599, 2022

    work page 2022

  35. [35]

    Causal inference, 2010

    Miguel Hern´ an and James Robins. Causal inference, 2010

    work page 2010

  36. [36]

    Bayesian nonparametric modeling for causal inference

    Jennifer Hill. Bayesian nonparametric modeling for causal inference. Journal of Computational and Graphical Statistics, 20(1):217–240, 2011

    work page 2011

  37. [37]

    Causal discovery from heterogeneous/nonstationary data

    Biwei Huang, Kun Zhang, Jiji Zhang, Joseph Ramsey, Ruben Sanchez-Romero, Clark Glymour, and Bernhard Sch¨ olkopf. Causal discovery from heterogeneous/nonstationary data. Journal of Machine Learning Research, 21(89):1–53, 2020

    work page 2020

  38. [38]

    Falsification before extrapolation in causal effect estimation

    Zeshan Hussain, Michael Oberst, Ming-Chieh Shih, and David Sontag. Falsification before extrapolation in causal effect estimation. Advances in Neural Information Processing Systems , 35, 2022

    work page 2022

  39. [39]

    Falsification of internal and external validity in observational studies via conditional moment restrictions

    Zeshan Hussain, Ming-Chieh Shih, Michael Oberst, Ilker Demirel, and David Sontag. Falsification of internal and external validity in observational studies via conditional moment restrictions. International Conference on Artificial Intelligence and Statistics , 2023

    work page 2023

  40. [40]

    Do-calculus when the true graph is unknown

    Antti Hyttinen, Frederick Eberhardt, and Matti J¨ arvisalo. Do-calculus when the true graph is unknown. Uncertainty in Artificial Intelligence , 2015

    work page 2015

  41. [41]

    Unpacking the black box of causality: Learning about causal mechanisms from experimental and observational studies

    Kosuke Imai, Luke Keele, Dustin Tingley, and Teppei Yamamoto. Unpacking the black box of causality: Learning about causal mechanisms from experimental and observational studies. American Political Science Review, 105(4):765–789, 2011

    work page 2011

  42. [42]

    Categorical reparameterization with gumbel-softmax

    Eric Jang, Shixiang Gu, and Ben Poole. Categorical reparameterization with gumbel-softmax. Inter- national Conference on Learning Representations, 2017

    work page 2017

  43. [43]

    Instrumental variables estimation with some invalid instruments and its application to mendelian randomization

    Hyunseung Kang, Anru Zhang, Tony Cai, and Dylan Small. Instrumental variables estimation with some invalid instruments and its application to mendelian randomization. Journal of the American Statistical Association, 111(513):132–144, 2016

    work page 2016

  44. [44]

    Demystifying double robustness: a comparison of alternative strategies for estimating a population mean from incomplete data

    Joseph Kang and Joseph Schafer. Demystifying double robustness: a comparison of alternative strategies for estimating a population mean from incomplete data. Statistical Science, pages 523–539, 2007

    work page 2007

  45. [45]

    Detecting hidden confounding in observational data using multiple environments

    Rickard Karlsson and Jesse Krijthe. Detecting hidden confounding in observational data using multiple environments. Advances in Neural Information Processing Systems , 37, 2023

    work page 2023

  46. [46]

    Dimension-agnostic inference using cross U-statistics

    Ilmun Kim and Aaditya Ramdas. Dimension-agnostic inference using cross U-statistics. Bernoulli, 30 (1):683–711, 2024

    work page 2024

  47. [47]

    A hard unsolved problem? Post-treatment bias in big social science questions

    Gary King. A hard unsolved problem? Post-treatment bias in big social science questions. Hard Problems in Social Science Symposium , 2010

    work page 2010

  48. [48]

    Ivy: Instrumental variable synthesis for causal inference

    Zhaobin Kuang, Frederic Sala, Nimit Sohoni, Sen Wu, Aldo C´ ordova-Palomera, Jared Dunnmon, James Priest, and Christopher R´ e. Ivy: Instrumental variable synthesis for causal inference. International Conference on Artificial Intelligence and Statistics , 2020. 14

    work page 2020

  49. [49]

    A generalized back-door criterion

    Marloes Maathuis and Diego Colombo. A generalized back-door criterion. The Annals of Statistics , 43 (3):1060–1088, 2015

    work page 2015

  50. [50]

    Estimating high-dimensional intervention effects from observational data

    Marloes Maathuis, Markus Kalisch, and Peter B¨ uhlmann. Estimating high-dimensional intervention effects from observational data. The Annals of Statistics , 37(6A):3133–3164, 2009

    work page 2009

  51. [51]

    The concrete distribution: A continuous relaxation of discrete random variables

    Chris Maddison, Andriy Mnih, and Yee Whye Teh. The concrete distribution: A continuous relaxation of discrete random variables. International Conference on Learning Representations, 2017

    work page 2017

  52. [52]

    Estimating bounds on causal effects in high-dimensional and possibly confounded systems

    Daniel Malinsky and Peter Spirtes. Estimating bounds on causal effects in high-dimensional and possibly confounded systems. International Journal of Approximate Reasoning , 88:371–384, 2017

    work page 2017

  53. [53]

    Identifying confounding from causal mechanism shifts

    Sarah Mameche, Jilles Vreeken, and David Kaltenpoth. Identifying confounding from causal mechanism shifts. In International Conference on Artificial Intelligence and Statistics , 2024

    work page 2024

  54. [54]

    Maternal cigarette smoking and perinatal mortality

    Mary Meyer and George Comstock. Maternal cigarette smoking and perinatal mortality. American Journal of Epidemiology , 96(1):1–10, 1972

    work page 1972

  55. [55]

    How conditioning on posttreatment variables can ruin your experiment and what to do about it.American Journal of Political Science, 62(3):760–775, 2018

    Jacob Montgomery, Brendan Nyhan, and Michelle Torres. How conditioning on posttreatment variables can ruin your experiment and what to do about it.American Journal of Political Science, 62(3):760–775, 2018

    work page 2018

  56. [56]

    A double machine learning approach to combining experimental and observational data

    Marco Morucci, Vittorio Orlandi, Harsh Parikh, Sudeepa Roy, Cynthia Rudin, and Alexander Volfovsky. A double machine learning approach to combining experimental and observational data. arXiv preprint arXiv:2307.01449, 2023

    work page arXiv 2023

  57. [57]

    Causal diagrams for empirical research

    Judea Pearl. Causal diagrams for empirical research. Biometrika, 82(4):669–688, 1995

    work page 1995

  58. [58]

    Direct and indirect effects

    Judea Pearl. Direct and indirect effects. Probabilistic and causal inference: the works of Judea Pearl , pages 373–392, 2022

    work page 2022

  59. [59]

    Interpreting and using CPDAGs with back- ground knowledge

    Emilija Perkovi´ c, Markus Kalisch, and Marloes Maathuis. Interpreting and using CPDAGs with back- ground knowledge. Uncertainty in Artificial Intelligence , 2017

    work page 2017

  60. [60]

    Complete graphical charac- terization and construction of adjustment sets in Markov equivalence classes of ancestral graphs.Journal of Machine Learning Research, 18(220):1–62, 2018

    Emilija Perkovi´ c, Johannes Textor, Markus Kalisch, and Marloes Maathuis. Complete graphical charac- terization and construction of adjustment sets in Markov equivalence classes of ancestral graphs.Journal of Machine Learning Research, 18(220):1–62, 2018

    work page 2018

  61. [61]

    Causal inference by using invariant prediction: identification and confidence intervals

    Jonas Peters, Peter B¨ uhlmann, and Nicolai Meinshausen. Causal inference by using invariant prediction: identification and confidence intervals. Journal of the Royal Statistical Society Series B: Statistical Methodology, 78(5):947–1012, 2016

    work page 2016

  62. [62]

    Elements of causal inference: foundations and learning algorithms

    Jonas Peters, Dominik Janzing, and Bernhard Sch¨ olkopf. Elements of causal inference: foundations and learning algorithms. The MIT Press , 2017

    work page 2017

  63. [63]

    Invariant causal prediction for sequential data

    Niklas Pfister, Peter B¨ uhlmann, and Jonas Peters. Invariant causal prediction for sequential data. Journal of the American Statistical Association , 2019

    work page 2019

  64. [64]

    Stabilizing variable selection and regression

    Niklas Pfister, Evan Williams, Jonas Peters, Ruedi Aebersold, and Peter B¨ uhlmann. Stabilizing variable selection and regression. The Annals of Applied Statistics , 15(3):1220–1246, 2021

    work page 2021

  65. [65]

    SPSS and SAS procedures for estimating indirect effects in simple mediation models

    Kristopher Preacher and Andrew Hayes. SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior research methods, instruments, & computers , 36:717–731, 2004

    work page 2004

  66. [66]

    Identifiability and exchangeability for direct and indirect effects

    James Robins and Sander Greenland. Identifiability and exchangeability for direct and indirect effects. Epidemiology, 3(2):143–155, 1992. 15

    work page 1992

  67. [67]

    Semiparametric efficiency in multivariate regression models with missing data

    James Robins and Andrea Rotnitzky. Semiparametric efficiency in multivariate regression models with missing data. Journal of the American Statistical Association , 90(429):122–129, 1995

    work page 1995

  68. [68]

    Estimation of regression coefficients when some regressors are not always observed

    James Robins, Andrea Rotnitzky, and Lue Ping Zhao. Estimation of regression coefficients when some regressors are not always observed. Journal of the American Statistical Association , 89(427):846–866, 1994

    work page 1994

  69. [69]

    Invariant models for causal transfer learning

    Mateo Rojas-Carulla, Bernhard Sch¨ olkopf, Richard Turner, and Jonas Peters. Invariant models for causal transfer learning. Journal of Machine Learning Research , 19(36):1–34, 2018

    work page 2018

  70. [70]

    The risks of invariant risk minimiza- tion

    Elan Rosenfeld, Pradeep Kumar Ravikumar, and Andrej Risteski. The risks of invariant risk minimiza- tion. International Conference on Learning Representations, 2021

    work page 2021

  71. [71]

    Causal Dantzig

    Dominik Rothenh¨ ausler, Peter B¨ uhlmann, and Nicolai Meinshausen. Causal Dantzig. The Annals of Statistics, 47(3):1688–1722, 2019

    work page 2019

  72. [72]

    Anchor regression: Heterogeneous data meet causality

    Dominik Rothenh¨ ausler, Nicolai Meinshausen, Peter B¨ uhlmann, and Jonas Peters. Anchor regression: Heterogeneous data meet causality. Journal of the Royal Statistical Society Series B: Statistical Method- ology, 83(2):215–246, 2021

    work page 2021

  73. [73]

    Efficient adjustment sets for population average causal treat- ment effect estimation in graphical models

    Andrea Rotnitzky and Ezequiel Smucler. Efficient adjustment sets for population average causal treat- ment effect estimation in graphical models. Journal of Machine Learning Research , 21(1):7642–7727, 2020

    work page 2020

  74. [74]

    A clinical trial of change in maternal smoking and its effect on birth weight

    Mary Sexton and Richard Hebel. A clinical trial of change in maternal smoking and its effect on birth weight. Journal of the American Medical Association , (7):911–915, 1984

    work page 1984

  75. [75]

    Finding valid adjustments under non- ignorability with minimal DAG knowledge

    Abhin Shah, Karthikeyan Shanmugam, and Kartik Ahuja. Finding valid adjustments under non- ignorability with minimal DAG knowledge. International Conference on Artificial Intelligence and Statistics, 2022

    work page 2022

  76. [76]

    Front-door adjustment beyond Markov equivalence with limited graph knowledge

    Abhin Shah, Karthikeyan Shanmugam, and Murat Kocaoglu. Front-door adjustment beyond Markov equivalence with limited graph knowledge. Advances in Neural Information Processing Systems , 2024

    work page 2024

  77. [77]

    Estimating individual treatment effect: generalization bounds and algorithms

    Uri Shalit, Fredrik Johansson, and David Sontag. Estimating individual treatment effect: generalization bounds and algorithms. International Conference on Machine Learning , 2017

    work page 2017

  78. [78]

    Causality-oriented robustness: exploiting general additive interventions

    Xinwei Shen, Peter B¨ uhlmann, and Armeen Taeb. Causality-oriented robustness: exploiting general additive interventions. arXiv preprint arXiv:2307.10299 , 2023

    work page arXiv 2023

  79. [79]

    Invariant representation learning for treatment effect estimation

    Claudia Shi, Victor Veitch, and David Blei. Invariant representation learning for treatment effect estimation. Uncertainty in Artificial Intelligence , 2021

    work page 2021

  80. [80]

    On the validity of covariate adjustment for estimating causal effects

    Ilya Shpitser, Tyler VanderWeele, and James Robins. On the validity of covariate adjustment for estimating causal effects. Uncertainty in Artificial Intelligence , 2010

    work page 2010

Showing first 80 references.