Pith. sign in

REVIEW 4 major objections 6 minor 3 cited by

Contrastive representations of high-dimensional, structured treatments

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read To estimate causal effects of high-dimensional treatments, discard non-causal latents.

desk verdict A promising contrastive recipe undermined by an internally inconsistent SCM; the bias theorem's proof violates the paper's own model. read the letter →

arxiv 2411.19245 v1 pith:JBGD3HJI submitted 2024-11-28 stat.ML cs.AIcs.LG

Oriol Corcoll Andreu , Athanasios Vlontzos , Michael O'Riordan , Ciaran M. Gilligan-Lee This is my paper · ORCID
classification stat.MLcs.AIcs.LG MSC 62D2068T05
keywords causalinferencehigh-dimensionaltreatmentscontrastivelearningtreatmentrepresentationconditionalaverageeffectbackdooradjustmentdisentanglementstructured
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper addresses causal effect estimation when treatments are high-dimensional structured objects such as text, video, or molecules. The observed treatment T is assumed to be a mixture of latent causal factors TC and non-causal factors TnC, with the outcome Y depending only on TC and covariates X. The paper shows that standard back-door adjustment using T directly is biased, even with all confounders observed, whenever TnC correlates with confounders. It proves that using a learned representation ψ(T) that contains no information about TnC is necessary and sufficient for unbiased estimation, and provides a contrastive algorithm that provably block-identifies TC. Empirical results on synthetic, molecule, and recommender datasets support the claim that contrastive representations yield lower bias than existing methods.

What carries the argument

The contrastive pair construction is the mechanism that carries the argument: positive pairs are samples with X ≈ X′ and Y ≈ Y′ but T ≠ T′, which forces TC to be approximately equal while TnC varies; negative pairs have X ≈ X′ but Y not close, forcing TC to differ. Under smooth invertible functions, this augmentation scheme block-identifies TC by appealing to an existing theorem on self-supervised contrastive learning that isolates content from style. The resulting representation ψ(T) is then used as the treatment in back-door adjustment, which Theorem 2 shows is unbiased exactly when ψ(T) discards TnC.

What would settle it

Construct a synthetic dataset with the same structure as Figure 1 but add a direct edge from T to Y, e.g., Y = f(TC, X) + γ·T + εY, and estimate the CATE using the contrastive representation ψ(T) after training on this data; if the resulting estimate is biased, the paper's claim that ψ(T) discards all non-causal information and preserves CATE is falsified.

Watch

Extended reading notes

Core claim

The central claim is that causal effect estimation from high-dimensional treatments is unbiased if and only if the treatment representation ψ(T) contains no information about non-causal latent variables TnC. Under structural equations X = l(εX), TC = g(X, εTC), TnC = h(X, εTnC), T = m(TC, TnC), and Y = f(TC, X, εY), the outcome depends on the treatment only through TC. The paper constructs a contrastive learning method that creates positive pairs from samples sharing the same X and Y (hence the same TC, since f is invertible) and negative pairs with different Y (hence different TC). Applying a known block-identification result, the method provably yields a representation ψ(T) = φ(TC) that isolates the causal latents. This representation preserves interventions on TC and consequently gives unbiased CATE estimates, whereas using T directly fails when TnC acts as a proxy for confounders.

Load-bearing premise

The outcome Y is assumed to depend on the treatment only through the latent causal components TC, not on the observed treatment T directly; if T itself has a direct causal effect on Y, the contrastive representation may not preserve the intervention and the unbiasedness guarantee collapses.

Editorial extensions

If this is right

  • Causal effect estimation for text, video, audio, and molecular treatments can be made unbiased by learning a contrastive representation that isolates the causally relevant latents.
  • Recommendation systems and drug discovery could benefit: if the causal aspects of a product or molecule are identified, similar causal profiles can be matched more efficiently.
  • The paper's Theorem 2 provides a necessary and sufficient condition for unbiasedness, giving practitioners a clear criterion for when a treatment representation is valid.
  • Existing methods for high-dimensional treatment effect estimation that use the treatment directly, such as SIN, are shown to be biased in this setting, motivating adoption of contrastive pre-training.
  • The fully non-parametric nature of the approach means it can be applied without specifying a parametric model of the treatment-outcome relationship.

Reading between the lines

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

  • If the outcome depends on T directly (violating Y = f(TC, X, εY)), the contrastive representation may not preserve the intervention on T, and the unbiasedness guarantee could fail; a modified estimand or additional assumptions would be needed.
  • The practical algorithm uses thresholds δ and ε to construct pairs; the bias introduced by these approximations remains unquantified, so a sensitivity analysis relative to δ and ε is a natural next step.
  • The block-identification result suggests the method could extend to settings with multiple treatment views or where partially observed causal latents exist, potentially enabling semi-supervised causal representation learning.
  • The authors' approach could be tested on text sentiment examples (e.g., tone vs. style) to verify empirically that the learned representation isolates tone and discards style, as the motivating example suggests.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper studies causal effect estimation when treatments are high-dimensional structured objects such as text, molecules, or product reviews. It models a treatment T as a function of causal latent factors TC and non-causal latent factors TnC, with outcome Y depending only on TC and covariates X. The paper claims that backdoor adjustment using T directly is biased when TnC is correlated with X, and it proposes a contrastive learning method to learn a representation ψ(T) that discards TnC. The theoretical contributions are stated as three theorems: Theorem 1 asserts bias of direct adjustment; Theorem 2 asserts that unbiased estimation is possible if and only if the representation contains no TnC information; and Theorem 3 asserts that the contrastive approach block-identifies TC. Experiments on synthetic, molecule, and recommender datasets compare the contrastive method to baselines using the PEHE metric.

Significance. The problem is practically relevant, and the idea of using contrastive learning to isolate the causal components of high-dimensional treatments is interesting and potentially useful. If the theoretical results were sound, the paper would offer a principled representation learning procedure for causal inference. However, the paper's own structural causal model renders the target estimand degenerate, and the proof of the central bias claim contradicts the model's structural equations. The empirical evaluation, while suggestive, is not anchored to a well-defined causal estimand and cannot rescue the theoretical failure. The manuscript therefore does not currently establish its central claims.

major comments (4)
  1. [Section 3 (structural equations) and Section 2 (Eq. (2))] Under the SCM in Section 3, with Y = f(TC, X, εY) and T = m(TC, TnC), intervening on T via do(T=t) replaces only the equation for T and leaves TC, X, and εY unchanged. Consequently E[Y | do(T=t), X=x] = E[Y | X=x] for all t, and the CATE τ(T,T',x) defined in Eq. (2) is identically zero. The paper's theorems discuss unbiasedness relative to the latent CATE τ(TC,T'C,X), which is a different estimand from the one introduced in Section 2. This is a fundamental mismatch: the paper's motivating problem, bias in the effect of the observed treatment T, is not well-defined under the proposed model.
  2. [Theorem 1 (proof)] The proof constructs an alternative outcome model Y = αTC + σ²βTnC + σ²X + εY, which makes Y depend directly on TnC and therefore violates the structural equation Y = f(TC, X, εY) stated at the start of Section 3. In the paper's own linear DGP, E[Y | T=(tc,tnc), X=x] = ρtc + δx, which is independent of tnc; hence backdoor adjustment with T correctly identifies the conditional average treatment effect of TC. The claimed counterexample does not occur under the stated model, and the proof's appeal to 'a possible solution to regressing Y on T and X' conflates the true conditional expectation with an arbitrary alternative model that is not compatible with the SCM.
  3. [Theorem 2] The statement is largely definitional. The paper defines unbiased CATE estimation as the condition that τ(T,T',X) reproduces τ(TC,T'C,X) in Section 3. Under that definition, the 'only if' direction essentially restates that a representation that reproduces the latent CATE cannot depend on TnC. The proof's key step—from equality of conditional expectations E[Y|ψ(T),X] = E[Y|ψ(T'),X] to the conclusion that ψ is constant on each TTC—assumes that ψ(T) itself is the quantity being intervened on, which is not established. The theorem therefore does not provide the claimed necessary-and-sufficient characterization, and it does not confront the degenerate estimand of Eq. (2).
  4. [Theorem 3 and Section 4 (algorithm)] The proof consists of an invocation of Theorem 4.2 of Von Kügelgen et al. (2021) rather than a derivation. The augmentation argument requires exact equality of X and Y values to conclude TC = T'C via invertibility of f; the practical relaxation with thresholds δ and ε in Section 4 is not proved to preserve block identification. As a result, the claim that the contrastive algorithm 'provably' block-identifies the causal latents is not supported for the actual algorithm evaluated in Section 6.
minor comments (6)
  1. [Introduction] In the Introduction, 'we cloud improve product recommendation' should be 'we could improve'; please correct this typo.
  2. [Section 4 and References] 'V on Kügelgen et al.' should be 'Von Kügelgen et al.'; the same typo appears in the References entry for Von Kügelgen et al. (2021).
  3. [Section 6, Eq. (1)] The PEHE formula in Eq. (1) omits the expectation or averaging convention; as written it is not clear whether the average is over a fixed test set or the population.
  4. [Figure 2b] The caption of Figure 2b, 'Performance under perturbations to y', does not match the text, which describes differences in predictions under perturbed non-causal treatment components.
  5. [Section 3] The notation f is used both for the outcome structural equation and for the treatment mixture T = f(TC,TnC) in Section 3; this is confusing and should be disambiguated.
  6. [Tables 1 and 2] The standard errors are formatted inconsistently (e.g., '0 .74' should be '0.74'); please correct the spacing throughout the tables.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation chain is either self-contained or imports an external (non-self-cited) block-identification theorem; the main issues are correctness gaps, not input–output equivalence.

full rationale

We walked the paper's derivation chain. Theorem 1 attempts to show bias of direct back-door adjustment; its proof constructs an alternative Y that depends on TnC, which violates the paper's own structural equation Y=f(TC,X,εY). This is a logical flaw in the counterexample, but it is not a circular reduction: the proof does not equate an output with an input, it uses an inadmissible model. Theorem 2's sufficiency direction is a near-restatement of the paper's definition of 'unbiased' (matching TC's CATE), but the necessity direction is an attempted independent argument; any issue there is a validity gap rather than an equivalence-by-construction. Theorem 3 relies on a theorem from Von Kügelgen et al. (2021), an external, non-self-cited source; under the stated smooth/invertibility assumptions, the positive/negative pair construction uses Y as supervision to hold TC fixed, which is a legitimate use of outcome data, not a fitted parameter renamed as a prediction. The empirical PEHE evaluation measures the same invariance that the contrastive loss is trained to produce, but the paper does not present this as a first-principles prediction; it is a direct validation of the training objective. No load-bearing self-citation chain appears: the many self-citations in the introduction are motivational only. We therefore find no significant circularity, while noting that the soundness of Theorem 1 and Theorem 2 is a separate correctness concern outside this rubric.

Assumptions & free parameters 5 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a specific structural causal model, strong invertibility assumptions, and an external identifiability theorem. The hand-chosen hyperparameters in the algorithm are not the main circularity driver, but the 'if and only if' theorem is close to a definitional restatement. The invented latent decomposition TC/TnC is assumed without independent evidence.

free parameters (5)
  • threshold_delta = not specified
    Algorithm 1 requires |g(x)-g(x')| ≤ δ for positive and negative pairs, but δ is never given a value or tuned in the paper.
  • threshold_epsilon = not specified
    Algorithm 1 requires |y-y'| ≤ ε for positive pairs and |y-y'| > ε for negative pairs, but ε is not specified.
  • triplet_loss_margin = 30 (Synthetic), 100 (Molecule and Recommender)
    Appendix B states the margin hyperparameter of the triplet loss is set to 30 for the Synthetic dataset and 100 for the real datasets, chosen by hand.
  • contrastive_loss_weight = 0.1 (Synthetic), 1.0 (Molecule and Recommender)
    Appendix B states the triplet loss is weighted by 0.1 for Synthetic and 1.0 for the other datasets, chosen per dataset.
  • model_coefficients_in_Theorem_1 = [α, σ^2β, σ^2]
    The counterexample model in Theorem 1 uses coefficients that are not derived from the true data generating process; they are chosen so that the model appears to match E[Y|TC,X], which is not actually the case.
assumptions (5)
  • domain assumption The outcome Y is generated by Y = f(TC, X, εY), so Y does not directly depend on the observed treatment T.
    Stated in the structural equations at the start of Section 3. This premise makes the estimand do(T) ill-defined because T is not a parent of Y in the DAG.
  • domain assumption All functions generating the latents and the outcome are smooth and invertible with smooth inverses, and noise distributions have positive density almost everywhere.
    Required by Theorem 3 for block-identification. This is a strong condition that is unlikely to hold for text, molecules, or other discrete structured treatments.
  • ad hoc to paper A representation ψ(T) that is constant on each set TTC is a reparametrization of TC.
    Used in the proof of Theorem 2. This requires a one-to-one correspondence between TC values and the sets {T = m(TC,TnC)}TnC, which is not established.
  • standard math Theorem 4.2 of Von Kügelgen et al. (2021) holds for the treatment setting.
    The identifiability proof of Theorem 3 is an application of this external theorem. The paper does not re-derive the theorem or verify all its conditions beyond citing it.
  • domain assumption There exist at least two different T values for the same (X,Y) values, so that TnC varies while TC is invariant.
    Invoked in the proof of Theorem 3 to guarantee that augmentations change TnC. This relies on non-trivial support of the noise term εTnC.
invented entities (1)
  • TC (causal latent factors) and TnC (non-causal latent factors)
    purpose: Latent decomposition of the high-dimensional treatment into outcome-relevant and outcome-irrelevant components.
    These latents are theoretical constructs assumed by the model. The paper provides no falsifiable handle to identify them in real data; the synthetic datasets encode them by construction, but the real-dataset evaluation does not validate the latent decomposition.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Contrastive representations of high-dimensional, structured treatments." pith.science (2026). https://pith.science/paper/JBGD3HJI

@misc{pith2026241119245,
  author       = {Pith},
  title        = {Pith review of: Contrastive representations of high-dimensional, structured treatments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JBGD3HJI}},
  note         = {Machine review of arXiv:2411.19245}
}
read the original abstract

Estimating causal effects is vital for decision making. In standard causal effect estimation, treatments are usually binary- or continuous-valued. However, in many important real-world settings, treatments can be structured, high-dimensional objects, such as text, video, or audio. This provides a challenge to traditional causal effect estimation. While leveraging the shared structure across different treatments can help generalize to unseen treatments at test time, we show in this paper that using such structure blindly can lead to biased causal effect estimation. We address this challenge by devising a novel contrastive approach to learn a representation of the high-dimensional treatments, and prove that it identifies underlying causal factors and discards non-causally relevant factors. We prove that this treatment representation leads to unbiased estimates of the causal effect, and empirically validate and benchmark our results on synthetic and real-world datasets.

Figures

Figures reproduced from arXiv: 2411.19245 by the authors.

Figure 1
Figure 1. DAG for our problem. While T depends on both TC and TnC in the structural equations, T = m(TC , TnC ) outcome Y only depends on T through TC : Y = f(TC , X, ϵY ). This is represented graphically by the blue arrow from TC to T, and on to Y , while the arrow from TnC to T is red. of them. That is, the treatment we’re given for a specific problem, T, corresponds to a (potentially non-linear) mixture of these latents, T… view at source ↗
Figure 3
Figure 3. Performance under different perturbations to the non-causal information of the treatment [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Entangled by Design: Spurious Intra-Variable Signal Routing in Tabular In-Context Learners

    cs.AI 2026-07 conditional novelty 5.0 of 10

    In-context learners route predictions through a spurious component inside a composite feature whenever that component correlates with the label, and the routing persists as context grows.

  2. Learning Treatment Representations for Downstream Instrumental Variable Regression

    cs.LG 2025-06 conditional novelty 5.0 of 10

    Instrument-guided representation learning, which folds instruments into the treatment encoder, yields representations on which IV regression identifies outcome-improving intervention directions.

  3. Agentic Personalisation of Cross-Channel Marketing Experiences

    cs.AI 2025-06 reject novelty 4.0 of 10

    An agentic marketing system combining Difference-in-Differences and Thompson sampling increased engagement and conversion in a 6.4-million-user field test, but the treatment was confounded with extra message volume.

Reference graph

Works this paper leans on

25 extracted references · 15 canonical work pages · cited by 3 Pith papers

  1. [1]

    Did i do that? blame as a means to identify controlled effects in reinforcement learning

    Oriol Corcoll, Youssef Sherif Mansour Mohamed, and Raul Vicente. Did i do that? blame as a means to identify controlled effects in reinforcement learning. Transactions on Machine Learning Research, 2022. ISSN 2835-8856. URL https://openreview.net/forum?id=NL2L3XjVFx

    work page 2022

  2. [2]

    Causing trouble

    Ciar \'a n Gilligan-Lee. Causing trouble. New Scientist, 246 0 (3279): 0 32--35, 2020

    work page 2020

  3. [3]

    Graphite: Estimating individual effects of graph-structured treatments

    Shonosuke Harada and Hisashi Kashima. Graphite: Estimating individual effects of graph-structured treatments. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management, pages 659--668, 2021

    work page 2021

  4. [4]

    Bayesian nonparametric modeling for causal inference

    Jennifer L Hill. Bayesian nonparametric modeling for causal inference. Journal of Computational and Graphical Statistics, 20 0 (1): 0 217--240, 2011

    2011

  5. [5]

    Disentangling causal effects from sets of interventions in the presence of unobserved confounders

    Olivier Jeunen, Ciar \'a n Gilligan-Lee, Rishabh Mehrotra, and Mounia Lalmas. Disentangling causal effects from sets of interventions in the presence of unobserved confounders. Advances in Neural Information Processing Systems, 35: 0 27850--27861, 2022

    work page 2022

  6. [6]

    Causal effect inference for structured treatments

    Jean Kaddour, Yuchen Zhu, Qi Liu, Matt J Kusner, and Ricardo Silva. Causal effect inference for structured treatments. Advances in Neural Information Processing Systems, 34: 0 24841--24854, 2021

    2021

  7. [7]

    Disentangled contrastive learning on graphs

    Haoyang Li, Xin Wang, Ziwei Zhang, Zehuan Yuan, Hang Li, and Wenwu Zhu. Disentangled contrastive learning on graphs. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan, editors, Advances in Neural Information Processing Systems, volume 34, pages 21872--21884. Curran Associates, Inc., 2021. URL https://proceedings.neurips.cc/pape...

    work page 2021

  8. [8]

    O’Neil, and Sotirios A

    Xiao Liu, Pedro Sanchez, Spyridon Thermos, Alison Q. O’Neil, and Sotirios A. Tsaftaris. Compositionally equivariant representation learning. IEEE Transactions on Medical Imaging, pages 1--1, 2024. doi:10.1109/TMI.2024.3358955

    arXiv 2024

Show all 25 references
  1. [9]

    Semiparametric causal sufficient dimension reduction of multidimensional treatments

    Razieh Nabi, Todd McNutt, and Ilya Shpitser. Semiparametric causal sufficient dimension reduction of multidimensional treatments. In Uncertainty in Artificial Intelligence, pages 1445--1455. PMLR, 2022

    2022

  2. [10]

    Representation learning with contrastive predictive coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Representation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748, 2018

    2018 arXiv

  3. [11]

    Spillover detection for donor selection in synthetic control models

    Michael O'Riordan and Ciar \'a n M Gilligan-Lee. Spillover detection for donor selection in synthetic control models. arXiv preprint arXiv:2406.11399, 2024

    2024 arXiv

  4. [12]

    Causality (2nd edition)

    Judea Pearl. Causality (2nd edition). Cambridge University Press, 2009

    2009

  5. [13]

    Causal effects of linguistic properties

    Reid Pryzant, Dallas Card, Dan Jurafsky, Victor Veitch, and Dhanya Sridhar. Causal effects of linguistic properties. arXiv preprint arXiv:2010.12919, 2020

    2010 arXiv

  6. [14]

    Quantum chemistry structures and properties of 134 kilo molecules

    Raghunathan Ramakrishnan, Pavlo O Dral, Matthias Rupp, and O Anatole Von Lilienfeld. Quantum chemistry structures and properties of 134 kilo molecules. Scientific data, 1 0 (1): 0 1--7, 2014

    2014

  7. [15]

    D’artagnan: Counterfactual video generation

    Hadrien Reynaud, Athanasios Vlontzos, Mischa Dombrowski, Ciar \'a n Gilligan Lee, Arian Beqiri, Paul Leeson, and Bernhard Kainz. D’artagnan: Counterfactual video generation. In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 599--6...

    2022

  8. [16]

    Improving the accuracy of medical diagnosis with causal machine learning

    Jonathan G Richens, Ciar \'a n M Lee, and Saurabh Johri. Improving the accuracy of medical diagnosis with causal machine learning. Nature communications, 11 0 (1): 0 3923, 2020

    2020

  9. [17]

    Recommendations as treatments: Debiasing learning and evaluation

    Tobias Schnabel, Adith Swaminathan, Ashudeep Singh, Navin Chandak, and Thorsten Joachims. Recommendations as treatments: Debiasing learning and evaluation. In Maria Florina Balcan and Kilian Q. Weinberger, editors, Proceedings of The 33rd International Conference on Machine Le...

    2016

  10. [18]

    Facenet: A unified embedding for face recognition and clustering

    Florian Schroff, Dmitry Kalenichenko, and James Philbin. Facenet: A unified embedding for face recognition and clustering. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 815--823, 2015

    2015

  11. [19]

    Contrastive embedding of structured space for bayesian optimization

    Josh Tingey, Ciar \'a n Mark Gilligan-Lee, and Zhenwen Dai. Contrastive embedding of structured space for bayesian optimization. In Fifth Workshop on Meta-Learning at the Conference on Neural Information Processing Systems, 2021

    2021

  12. [20]

    Estimating long-term causal effects from short-term experiments and long-term observational data with unobserved confounding

    Graham Van Goffrier, Lucas Maystre, and Ciar \'a n Mark Gilligan-Lee. Estimating long-term causal effects from short-term experiments and long-term observational data with unobserved confounding. In Conference on Causal Learning and Reasoning, pages 791--813. PMLR, 2023

    2023

  13. [21]

    Estimating categorical counterfactuals via deep twin networks

    Athanasios Vlontzos, Bernhard Kainz, and Ciar \'a n M Gilligan-Lee. Estimating categorical counterfactuals via deep twin networks. Nature Machine Intelligence, 5 0 (2): 0 159--168, 2023

    2023

  14. [22]

    u gelgen, Yash Sharma, Luigi Gresele, Wieland Brendel, Bernhard Sch \

    Julius Von K \"u gelgen, Yash Sharma, Luigi Gresele, Wieland Brendel, Bernhard Sch \"o lkopf, Michel Besserve, and Francesco Locatello. Self-supervised learning with data augmentations provably isolates content from style. Advances in neural information processing systems, 34:...

    2021

  15. [23]

    Self-supervised learning disentangled group representation as feature

    Tan Wang, Zhongqi Yue, Jianqiang Huang, Qianru Sun, and Hanwang Zhang. Self-supervised learning disentangled group representation as feature. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan, editors, Advances in Neural Information Processing Syste...

    2021

  16. [24]

    The cancer genome atlas pan-cancer analysis project

    John N Weinstein, Eric A Collisson, Gordon B Mills, Kenna R Shaw, Brad A Ozenberger, Kyle Ellrott, Ilya Shmulevich, Chris Sander, and Joshua M Stuart. The cancer genome atlas pan-cancer analysis project. Nature genetics, 45 0 (10): 0 1113--1120, 2013

    2013

  17. [25]

    Non-parametric identifiability and sensitivity analysis of synthetic control models

    Jakob Zeitler, Athanasios Vlontzos, and Ciar \'a n Mark Gilligan-Lee. Non-parametric identifiability and sensitivity analysis of synthetic control models. In Conference on Causal Learning and Reasoning, pages 850--865. PMLR, 2023

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.