A Partial Initialization Strategy to Mitigate the Overfitting Problem in CATE Estimation with Hidden Confounding

Chuan Zhou; Chunyuan Zheng; Haiteng Zhang; Haoxuan Li; Mingming Gong; Yaxuan Li

REVIEW 4 major objections 6 minor 52 references

A Partial Initialization Strategy to Mitigate the Overfitting Problem in CATE Estimation with Hidden Confounding

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Two-stage pretraining-finetuning with partial head initialization avoids small-RCT overfitting and delivers competitive CATE estimates under hidden confounding.

desk verdict Real empirical wins, but the paper never isolates the partial initialization it credits; a duplicated baseline row and a randomized validation set further undercut the central claim. read the letter →

arxiv 2501.08888 v2 pith:OFAPG3QB submitted 2025-01-15 cs.LG

Chuan Zhou , Yaxuan Li , Chunyuan Zheng , Haiteng Zhang , Haoxuan Li , Mingming Gong This is my paper · ORCID

classification cs.LG MSC 62D2068T07

keywords CATEestimationhiddenconfoundingobservationalandRCTdatafusiontwo-stagepretraining-finetuningpartialparameterinitializationrepresentationlearningmutualinformationregularizationoverfittingmitigation

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

Estimating conditional average treatment effects (CATE) from observational data is threatened by unmeasured confounders, while randomized controlled trials are unconfounded but often too small to train neural predictors. This paper tries to establish that a two-stage pretraining-finetuning strategy solves that tension: first fit a covariate representation and outcome heads on large observational data, then finetune on small RCT data using an adapter representation and prediction heads partially initialized from the first stage. The proposed TSPF framework is claimed to avoid overfitting from the small RCT sample and to produce the most competitive CATE estimates among the compared methods on the IHDP and Jobs benchmarks. If correct, this would make it practical to combine abundant biased data with scarce unbiased data for causal effect estimation without imposing linear or additive data-generation assumptions.

What carries the argument

The load-bearing mechanism is the partial initialization of the second-stage heads, supported by the representation adapter and the shift regularizer. Because the new head $g_t$ has wider layers than the pretrained $h_t$ to ingest the concatenated representation $[Z; Z_U]$, its parameters are partitioned into a shared block (same shape as $h_t$) and new blocks; shared weights and biases are copied from $h_t$ and every new block is zeroed. At initialization, the augmented branch therefore contributes nothing and the model's predictions coincide with the first-stage model, so finetuning on RCT data starts from the pretrained solution rather than from noise. The $\mathcal{L}_{\mathrm{shift}}$ penalty then constrains $g_0,g_1$ to stay close to that starting point, and the CLUB mutual-information term prevents the adapter representation $Z_U$ from simply copying $Z$. Together these pieces are intended to let the RCT data remove the hidden-confounding bias while preserving the covariate structure learned from the large observational dataset.

What would settle it

A direct ablation that resets the shared blocks of $g_0$ and $g_1$ to random values while keeping the adapter, the shift regularizer, the loss weights, and the optimizer identical would settle the claim: if the random-initialized version matches or beats TSPF's out-of-sample $\sqrt{\epsilon_{\mathrm{PEHE}}}$ on IHDP or Jobs, then the partial initialization is not the source of the overfitting reduction.

Watch

Extended reading notes

Core claim

The central claim is that the partial initialization of the second-stage prediction heads is what lets the model learn an unbiased CATE from small RCT data without overfitting. In the first stage, the representation module and heads $h_0,h_1$ are trained on observational data, which contain hidden confounding and therefore give biased outcomes. In the second stage, the first-stage representation is frozen; an adapter module produces an augmented representation $Z_U$, the two are concatenated, and new heads $g_0,g_1$ are finetuned on RCT data. The shared block of each $g_t$ weight matrix is initialized to $h_t$'s weights, the new rows and columns are set to zero, and the bias vectors are initialized similarly, so the finetuned model starts by reproducing the first-stage predictions. A shift regularizer keeps the finetuned parameters near this initialization, and a contrastive mutual-information bound keeps $Z_U$ from duplicating $Z$. The paper reports that this design outperforms the compared baselines, markedly beating the related two-stage method CorNet on both datasets.

Load-bearing premise

The claim rests on the assumption that prediction heads pretrained on biased observational outcomes are a better starting point for learning the true CATE from RCT data than random initialization, and that the shift penalty keeping them close to that biased starting point will not simply preserve the confounding bias.

Editorial extensions

If this is right

If the claim holds, CATE models can exploit large biased observational datasets plus tiny RCT samples without needing untestable assumptions such as instrumental variables or front-door adjustment.
Finetuning only the initialized heads and a small adapter means the effective number of parameters trained on RCT data is small, which is precisely the mechanism claimed to curb overfitting.
The adapter's width and depth can be shrunk or grown with the RCT sample size, giving a practical lever for overfitting control.
The method's advantage over CorNet suggests that learning a residual representation, rather than only estimating a residual function, is a productive direction for data fusion in causal inference.

Reading between the lines

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

Beyond the paper, a direct ablation holding the adapter, loss weights, and optimizer fixed while resetting the shared head blocks to random values would isolate whether the partial initialization itself, rather than the architecture or the shift regularizer, is responsible for the reported gains.
Beyond the paper, the shift regularizer may still bias the final estimate toward the observational solution, so an open scaling question is how much RCT data is needed to overcome that bias as confounding strength grows.
Beyond the paper, the same two-stage recipe should transfer to binary or survival outcomes, since the first-stage prediction loss already supports cross-entropy and the adapter width is tunable.

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.

Desk Editor's Note

read the letter

Quick take: the paper's empirical wins are real, but the claimed mechanism—partial initialization—is never isolated, and the evaluation protocol has a couple of red flags. If you cite it, cite the configuration, not the overfitting story.

What's actually new: a two-stage pretraining-finetuning setup for CATE with hidden confounding. Stage one trains a representation and two heads on large OBS data. Stage two freezes that representation, adds a smaller adapter representation, concatenates them, widens the heads, copies the pretrained weights into the shared part, and zero-initializes the new rows. On top of that it adds a CLUB mutual-information penalty between the two representations and an L2 penalty on how far the heads move from their initialization. That combination doesn't appear in the cited Kallus et al. or CorNet work, and the empirical results on IHDP and Jobs are strong—TSPF is clearly better than the baselines on most metrics, and the gap over CorNet is large.

What's soft. First, the central claim is about the partial initialization mitigating overfitting from small RCT data, but no ablation isolates it. Table 1 compares complete frameworks; the gains could come from the frozen representation plus adapter, from the CLUB term, from the L_shift regularizer, or from the initialization. The paper says the initialization is crucial but gives no experiment that varies it. Second, the data-generating process in Section 4.2 is linear: every term is a dot product W·x or W·u plus Gaussian noise. The paper claims generality to nonlinear mechanisms, but the experiments don't exercise any. Third, the validation set is converted to a randomized distribution by reassigning treatment with a fair coin and swapping in the counterfactual outcome. That means out-of-sample results are on a distribution with no confounding, not the observational distribution the method is supposed to correct. Fourth, in Table 1 the CorNet row is identical to the DRCFR row on every single metric. Either they didn't actually run CorNet and copied, or there's a reporting error. That alone makes the experimental table unusable as a comparison to the key two-stage baseline.

Bottom line: the idea is worth discussing and the configuration is new, but the evidence for the proposed mechanism is not there yet. A serious referee should ask for an ablation of the initialization, a fix for the CorNet/DRCFR duplicate, and an evaluation that respects the observational validation distribution. I'd send it to review—the core question is answerable with a few experiments—but not accept it as-is.

Referee Report

4 major / 6 minor

Summary. The paper proposes a two-stage pretraining-finetuning framework (TSPF) for estimating conditional average treatment effects when observational data are confounded by unmeasured variables and only a small randomized controlled trial (RCT) is available. In the first stage, a representation network and two outcome heads are trained on large observational data. In the second stage, the representation is frozen, a small adapter network is introduced, the representation and adapter outputs are concatenated, and the two outcome heads are finetuned on the RCT data with a partial initialization taken from the first-stage heads. The method also uses a CLUB mutual-information penalty between the two representations and an L_shift regularizer that anchors the finetuned heads to their initialized values. Experiments on two semi-synthetic datasets (IHDP and Jobs) report lower PEHE and ATE errors than a set of baselines, and the paper claims that the partial initialization mitigates overfitting from the small RCT sample.

Significance. If the central claim holds, the paper addresses a practical and important problem: combining biased large-scale observational data with small unconfounded RCT data for heterogeneous treatment effect estimation. The two-stage design is intuitive, and the partial initialization idea is a reasonable mechanism to combat overfitting in small RCT finetuning. The experimental comparison on two standard benchmarks is a useful empirical contribution, and the paper explicitly discusses the most related residual-correction baseline. However, the paper's own claim about the mechanism is not isolated by any ablation, and the synthetic experiments are confined to a linear and additive data-generating process, which undercuts the paper's stated generality. The significance of the work therefore depends on additional experiments that the authors have not reported.

major comments (4)

[Section 4.4, Table 1] The manuscript's central claim is that the partial initialization strategy in the second stage 'can avoid overfitting' caused by small RCT data. Table 1 only compares complete methods (TSPF versus baselines); no experiment varies the initialization scheme while holding the frozen representation φ, the adapter φ_U, the CLUB mutual-information penalty, and the L_shift regularization fixed. Without such an ablation, the observed gains cannot be attributed to the partial initialization; they could come from the frozen representation with a small adapter, from the mutual-information term, or from L_shift. Please add ablation studies that at least compare (a) partial initialization versus random initialization, (b) L_shift on versus off, and (c) combinations, with the same architecture and data.
[Section 4.2] The synthetic data-generating process is linear and additive: μ_0 = W_3·X + W_4·U, μ_1 = W_5·X + W_6·U + 4, with Gaussian noise and a logistic propensity. The contributions and conclusion claim that 'the proposed framework does not rely on the linear and additive generation assumptions' and that the method has the advantage of 'not restricting the data-generating process.' This claim is not supported by the experiments, which only exercise this linear setting. Please add nonlinear synthetic experiments (e.g., neural or polynomial outcome and propensity functions) or revise the claim to reflect the tested scope.
[Section 3.2, Regularization] The L_shift penalty is defined relative to the first-stage initialization θ_g^0, which comes from a model trained on confounded observational data. If λ_4 is large, the finetuned heads remain close to that biased initialization and the RCT signal may be unable to remove the hidden confounding. The paper does not report sensitivity to λ_4, nor the final distance ||θ_g − θ_g^0|| after training, so it is unclear whether the regularizer is actually preventing overfitting or inadvertently preserving confounding bias. Please add a sensitivity analysis over λ_4 and, if possible, an analysis of the parameter displacement.
[Section 4.2] The validation set is transformed by replacing the original treatment and factual outcome with an RCT-like assignment before evaluation. If the validation set is used for early stopping or hyperparameter selection, model selection is performed under an unconfounded assignment rather than the original observational distribution. The paper should clarify how the validation set was used and report out-of-sample performance under the original observational assignment, or explain why the replacement does not affect model selection.

minor comments (6)

[Table 1] The row label 'TSFP (ours)' should be 'TSPF (ours)'.
[Section 3.2.2] The activation function σ is used throughout the network definitions but is never specified; please state which activation function is used.
[Section 4.3.1] Kallus et al. [23] is discussed as the closest residual-correction baseline but is not included in the experiments; please add it or justify its omission.
[Section 4.3.3] Hyperparameter tuning is described only as 'tune the scale parameters from 1e−5 to 0.1'; provide the full search grid and selected values for reproducibility.
[Section 4.4] The phrase 'significantly outperforms' is used without any statistical significance tests; report paired tests or confidence intervals across the repeated runs.
[Section 4.2] The hidden confounder dimension is fixed at c = 30 and the RCT sample size is not varied; a sensitivity analysis over c and RCT sample size would strengthen the claim about overfitting mitigation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the TSPF framework is self-contained, and the missing ablation is an evidence gap, not a logical circle.

full rationale

The paper does not derive its central claim from a premise that already contains the conclusion. The first-stage objective (L_f + lambda1 * L_rec + lambda2 * L_unb) and the second-stage objective (L_pred + lambda3 * L_MI + lambda4 * L_shift) are ordinary supervised and regularized losses, and the reported CATE estimate is the difference of the two prediction-head outputs after RCT finetuning, so it is not forced by construction to equal the OBS-only estimate. The partial initialization rule (W_g, b_g) <- (W_h, 0, b_h, 0) is described as making the model "initially produce the same prediction as the trained first-stage model," but this is only an initialization condition; the second-stage parameters are subsequently updated with L_pred, and L_shift only penalizes deviation from the initial values. That is a modeling choice, not an identity between the final output and the input. No fitted parameter is renamed as a prediction, no uniqueness theorem from the authors' prior work is invoked, and the cited works by the authors appear as background rather than as load-bearing justification. The main weakness is that no ablation varies the initialization while holding L_shift, the frozen representation, the adapter, and the mutual-information term fixed, so Table 1 cannot isolate the mechanism claimed in Section 3.2. That is a missing-experiment concern and does not amount to circularity. The derivation chain is therefore self-contained; the appropriate finding is no significant circularity.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the causal identification assumptions for RCT data, the same-population assumption, and the untested heuristic that a small adapter plus a proximity penalty to a biased initialization can remove hidden confounding. The hyperparameters are tuned but not reported, and the data-generation setup fixes the confounder dimension to 30.

free parameters (6)

lambda_1 (reconstruction loss weight) = not reported (tuned from 1e-5 to 0.1)
Hyperparameter in first-stage objective tuned on validation.
lambda_2 (IPM loss weight) = not reported (tuned from 1e-5 to 0.1)
Hyperparameter in first-stage objective tuned on validation.
lambda_3 (CLUB MI loss weight) = not reported (tuned from 1e-5 to 0.1)
Hyperparameter in second-stage objective tuned on validation.
lambda_4 (L_shift regularization weight) = not reported (tuned from 1e-5 to 0.1)
Hyperparameter in second-stage objective tuned on validation.
adapter network width/depth = not reported
The paper says the adapter 'can be adjusted based on the scale of RCT data size' but does not report the chosen architecture.
c (hidden confounder dimension) = 30
Strength of simulated hidden confounding; a data-generation choice, not a model parameter.

assumptions (6)

domain assumption SUTVA (no interference, single version of treatment)
Stated in Section 2.2 as required for identification.
domain assumption RCT ignorability, consistency, positivity
Assumptions 1-3 in Section 2.2; needed to identify CATE from RCT data.
domain assumption OBS and RCT samples come from the same target population
Stated at the start of Section 2.1; if the distributions differ, the finetuning on RCT cannot correct OBS bias.
standard math CLUB provides a valid upper bound of mutual information given variational q_theta
Cites Cheng et al. [5]; used in L_MI in Section 3.2.1.
ad hoc to paper The representation adapter phi_U can capture the hidden confounding information missing from phi
Implicit in the method; no theory or experiment demonstrates that the adapter identifies the confounder direction.
ad hoc to paper L_shift regularization prevents overfitting without reintroducing confounding bias
The proximity penalty to a biased initialization could preserve bias; this is not analyzed.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Partial Initialization Strategy to Mitigate the Overfitting Problem in CATE Estimation with Hidden Confounding." pith.science (2026). https://pith.science/paper/OFAPG3QB

@misc{pith2026250108888,
  author       = {Pith},
  title        = {Pith review of: A Partial Initialization Strategy to Mitigate the Overfitting Problem in CATE Estimation with Hidden Confounding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OFAPG3QB}},
  note         = {Machine review of arXiv:2501.08888}
}

read the original abstract

Estimating the conditional average treatment effect (CATE) from observational data plays a crucial role in areas such as e-commerce, healthcare, and economics. Existing studies mainly rely on the strong ignorability assumption that there are no hidden confounders, whose existence cannot be tested from observational data and can invalidate any causal conclusion. In contrast, data collected from randomized controlled trials (RCT) do not suffer from confounding but are usually limited by a small sample size. To avoid overfitting caused by the small-scale RCT data, we propose a novel two-stage pretraining-finetuning (TSPF) framework with a partial parameter initialization strategy to estimate the CATE in the presence of hidden confounding. In the first stage, a foundational representation of covariates is trained to estimate counterfactual outcomes through large-scale observational data. In the second stage, we propose to train an augmented representation of the covariates, which is concatenated with the foundational representation obtained in the first stage to adjust for the hidden confounding. Rather than training a separate network from scratch, part of the prediction heads are initialized from the first stage. The superiority of our approach is validated on two datasets with extensive experiments.

Figures

Figures reproduced from arXiv: 2501.08888 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 45 canonical work pages

[1]

Susan Athey and Guido W Imbens. 2015. Machine learning methods for estimat- ing heterogeneous causal effects. Stat 1050, 5 (2015), 1–26

work page 2015
[2]

Florent Bédécarrats, Isabelle Guérin, and François Roubaud. 2020. Randomized control trials in the field of development: A critical perspective . Oxford University Press

work page 2020
[3]

Diogo GC Britto, Paolo Pinotti, and Breno Sampaio. 2022. The effect of job loss and unemployment insurance on crime in Brazil. Econometrica 90, 4 (2022), 1393–1423

work page 2022
[4]

Marie-Laure Charpignon, Bella Vakulenko-Lagun, Bang Zheng, Colin Magdamo, Bowen Su, Kyle Evans, Steve Rodriguez, Artem Sokolov, Sarah Boswell, Yi-Han Sheu, et al. 2022. Causal inference in medical records and complementary sys- tems pharmacology for metformin drug repurposing towards dementia. Nature Communications 13, 1 (2022), 7652

work page 2022
[5]

Pengyu Cheng, Weituo Hao, Shuyang Dai, Jiachang Liu, Zhe Gan, and Lawrence Carin. 2020. Club: A contrastive log-ratio upper bound of mutual information. In International Conference on Machine Learning . PMLR, 1779–1788

2020
[6]

Bénédicte Colnet, Imke Mayer, Guanhua Chen, Awa Dieng, Ruohong Li, Gaël Varoquaux, Jean-Philippe Vert, Julie Josse, and Shu Yang. 2024. Causal inference methods for combining randomized trials and observational studies: a review. Statist. Sci. 39, 1 (2024), 165–191

work page 2024
[7]

Giovanni Corrao, Federica Nicotra, Andrea Parodi, Antonella Zambon, Davide Soranna, Franca Heiman, Luca Merlino, and Giuseppe Mancia. 2012. External adjustment for unmeasured confounders improved drug–outcome association estimates based on health care utilization data. Journal of Clinical Epidemiology 65, 11 (2012), 1190–1199

work page 2012
[8]

Jacob Dorn, Kevin Guo, and Nathan Kallus. 2024. Doubly-valid/doubly-sharp sensitivity analysis for causal inference with unmeasured confounding. J. Amer. Statist. Assoc. (2024), 1–12

work page 2024

Show all 52 references

[9]

Charlie Frogner, Chiyuan Zhang, Hossein Mobahi, Mauricio Araya, and Tomaso A Poggio. 2015. Learning with a Wasserstein loss. Advances in Neural Information Processing Systems 28 (2015)

2015
[10]

Isabel R Fulcher, Ilya Shpitser, Stella Marealle, and Eric J Tchetgen Tchetgen. 2020. Robust inference on population indirect causal effects: the generalized front door criterion. Journal of the Royal Statistical Society Series B: Statistical Methodology 82, 1 (2020), 199–214

2020
[11]

Fernando Pires Hartwig, Linbo Wang, George Davey Smith, and Neil Martin Davies. 2023. Average causal effect estimation via instrumental variables: the no simultaneous heterogeneity assumption. Epidemiology 34, 3 (2023), 325–332

2023
[12]

Negar Hassanpour and Russell Greiner. 2019. Learning disentangled represen- tations for counterfactual regression. In International Conference on Learning Representations

2019
[13]

Tobias Hatt, Jeroen Berrevoets, Alicia Curth, Stefan Feuerriegel, and Mihaela van der Schaar. 2022. Combining observational and randomized data for estimat- ing heterogeneous treatment effects. arXiv:2202.12891 (2022)

2022 arXiv
[14]

Tobias Hatt, Daniel Tschernutter, and Stefan Feuerriegel. 2022. Generalizing off- policy learning under sample selection bias. InUncertainty in Artificial Intelligence. PMLR, 769–779

2022
[15]

Simon Haykin. 1994. Neural Networks: A Comprehensive Foundation . Macmillan

1994
[16]

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

2011
[17]

Harrell Jr, Karel GM Moons, Thomas PA Debray, and Johannes B Reitsma

Jeroen Hoogland, Joanna IntHout, Michail Belias, Maroeska M Rovers, Richard D Riley, Frank E. Harrell Jr, Karel GM Moons, Thomas PA Debray, and Johannes B Reitsma. 2021. A tutorial on individualized treatment effect prediction from randomized trials with a binary endpoint. Sta...

2021
[18]

Van-Nam Huynh, Vladik Kreinovich, and Songsak Sriboonchitta. 2016. Causal inference in econometrics. Springer

2016
[19]

Guido W Imbens. 2003. Sensitivity to exogeneity assumptions in program evalu- ation. American Economic Review 93, 2 (2003), 126–132

2003
[20]

Guido W Imbens and Donald B Rubin. 2015. Causal inference in statistics, social, and biomedical sciences. Cambridge university press

2015
[21]

Fredrik Johansson, Uri Shalit, and David Sontag. 2016. Learning representations for counterfactual inference. In International Conference on Machine Learning . PMLR, 3020–3029

2016
[22]

Imbens Joshua D

Guido W. Imbens Joshua D. Angrist and Donald B. Rubin. 1996. Identification of Causal Effects Using Instrumental Variables. J. Amer. Statist. Assoc. 91, 434 (1996), 444–455

1996
[23]

Nathan Kallus, Aahlad Manas Puli, and Uri Shalit. 2018. Removing hidden con- founding by experimental grounding. Advances in Neural Information Processing Systems 31 (2018)

2018
[24]

Edward H. Kennedy. 2023. Towards optimal doubly robust estimation of hetero- geneous causal effects. Electronic Journal of Statistics 17, 2 (2023), 3008 – 3049. https://doi.org/10.1214/23-EJS2157

2023 doi
[25]

Lingjing Kong, Guangyi Chen, Petar Stojanov, Haoxuan Li, Eric P Xing, and Kun Zhang. 2025. Towards Understanding Extrapolation: a Causal Lens. Advances in Neural Information Processing Systems 37 (2025)

2025
[26]

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

2019
[27]

Baohong Li, Haoxuan Li, Anpeng Wu, Minqin Zhu, Qingyu Cao, and Kun Kuang
[28]

Baohong Li, Haoxuan Li, Ruoxuan Xiong, Anpeng Wu, Fei Wu, and Kun Kuang
[29]

Haoxuan Li, Kunhan Wu, Chunyuan Zheng, Yanghao Xiao, Hao Wang, Zhi Geng, Fuli Feng, Xiangnan He, and Peng Wu. 2024. Removing hidden confounding in recommendation: a unified multi-task learning approach. Advances in Neural Information Processing Systems 36 (2024)

2024
[30]

In International Conference on Machine Learning

Learning Shadow Variable Representation for Treatment Effect Estimation under Collider Bias. In International Conference on Machine Learning . PMLR
[31]

Haoxuan Li, Chunyuan Zheng, Yixiao Cao, Zhi Geng, Yue Liu, and Peng Wu
[32]

Haoxuan Li, Yanghao Xiao, Chunyuan Zheng, and Peng Wu. 2023. Balancing un- observed confounding with a few unbiased ratings in debiased recommendations. In Proceedings of the ACM Web Conference 2023 . 1305–1313

2023
[33]

Christos Louizos, Uri Shalit, Joris M Mooij, David Sontag, Richard Zemel, and Max Welling. 2017. Causal effect inference with deep latent-variable models. Advances in Neural Information Processing Systems 30 (2017)

2017
[34]

Judea Pearl. 2009. Causality. Cambridge university press

2009
[35]

Xitong Li, Jörn Grahl, and Oliver Hinz. 2022. How do recommender systems lead to consumer purchases? A causal mediation analysis of a field experiment. Information Systems Research 33, 2 (2022), 620–637

2022
[36]

JAMES M Robins and MA Hernán. 2016. Causal inference

2016
[37]

Uri Shalit, Fredrik D Johansson, and David Sontag. 2017. Estimating individual treatment effect: generalization bounds and algorithms. In International Confer- ence on Machine Learning . PMLR, 3076–3085

2017
[38]

Mattia Prosperi, Yi Guo, Matt Sperrin, James S Koopman, Jae S Min, Xing He, Shannan Rich, Mo Wang, Iain E Buchan, and Jiang Bian. 2020. Causal inference and counterfactual prediction in machine learning for actionable healthcare. Nature Machine Intelligence 2, 7 (2020), 369–375

2020
[39]

Stefan Wager and Susan Athey. 2018. Estimation and inference of heterogeneous treatment effects using random forests. J. Amer. Statist. Assoc. 113, 523 (2018), 1228–1242

2018
[40]

Hao Wang, Jiajun Fan, Zhichao Chen, Haoxuan Li, Weiming Liu, Tianqiao Liu, Quanyu Dai, Yichao Wang, Zhenhua Dong, and Ruiming Tang. 2024. Opti- mal transport for treatment effect estimation. Advances in Neural Information Processing Systems 36 (2024)

2024
[41]

Claudia Shi, David Blei, and Victor Veitch. 2019. Adapting neural networks for the estimation of treatment effects. Advances in neural information processing systems 32 (2019)

2019
[42]

Peng Wu, Haoxuan Li, Yuhao Deng, Wenjie Hu, Quanyu Dai, Zhenhua Dong, Jie Sun, Rui Zhang, and Xiao-Hua Zhou. 2022. On the Opportunity of Causal Learning in Recommendation Systems: Foundation, Estimation, Prediction and Challenges. In International Joint Conference on Artificia...

2022
[43]

Yunfu Xu and Aiya Li. 2020. The relationship between innovative human capital and interprovincial economic growth based on panel data model and spatial econometrics. J. Comput. Appl. Math. 365 (2020), 112381

2020
[44]

Lili Wu and Shu Yang. 2022. Integrative𝑅-learner of heterogeneous treatment effects combining experimental and observational studies. InConference on Causal Learning and Reasoning. PMLR, 904–926

2022
[45]

Jinsung Yoon, James Jordon, and Mihaela Van Der Schaar. 2018. GANITE: Esti- mation of individualized treatment effects using generative adversarial nets. In International Conference on Learning Representations

2018
[46]

Emily C Zabor, Alexander M Kaizer, and Brian P Hobbs. 2020. Randomized controlled trials. Chest 158, 1 (2020), S79–S87

2020
[47]

Liuyi Yao, Sheng Li, Yaliang Li, Mengdi Huai, Jing Gao, and Aidong Zhang. 2018. Representation learning for treatment effect estimation from observational data. Advances in Neural Information Processing Systems 31 (2018)

2018
[48]

Kailiang Zhong, Fengtong Xiao, Yan Ren, Yaorong Liang, Wenqing Yao, Xiaofeng Yang, and Ling Cen. 2022. Descn: Deep entire space cross networks for individual treatment effect estimation. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . ...

2022
[49]

Minqin Zhu, Anpeng Wu, Haoxuan Li, Ruoxuan Xiong, Bo Li, Xiaoqing Yang, Xuan Qin, Peng Zhen, Jiecheng Guo, Fei Wu, and Kun Kuang. 2024. Contrastive balancing representation learning for heterogeneous dose-response curves esti- mation. In Proceedings of the AAAI Conference on A...

2024
[50]

Yao Zhang, Alexis Bellot, and Mihaela Schaar. 2020. Learning overlapping repre- sentations for the estimation of individualized treatment effects. In International Conference on Artificial Intelligence and Statistics . PMLR, 1005–1014

2020
[2023]

In International Conference on Machine Learning

Trustworthy policy learning under the counterfactual no-harm criterion. In International Conference on Machine Learning . PMLR, 20575–20598
[2024]

In International Conference on Machine Learning

A Generative Approach for Treatment Effect Estimation under Collider Bias: From an Out-of-Distribution Perspective. In International Conference on Machine Learning. PMLR

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Susan Athey and Guido W Imbens. 2015. Machine learning methods for estimat- ing heterogeneous causal effects. Stat 1050, 5 (2015), 1–26

work page 2015

[2] [2]

Florent Bédécarrats, Isabelle Guérin, and François Roubaud. 2020. Randomized control trials in the field of development: A critical perspective . Oxford University Press

work page 2020

[3] [3]

Diogo GC Britto, Paolo Pinotti, and Breno Sampaio. 2022. The effect of job loss and unemployment insurance on crime in Brazil. Econometrica 90, 4 (2022), 1393–1423

work page 2022

[4] [4]

Marie-Laure Charpignon, Bella Vakulenko-Lagun, Bang Zheng, Colin Magdamo, Bowen Su, Kyle Evans, Steve Rodriguez, Artem Sokolov, Sarah Boswell, Yi-Han Sheu, et al. 2022. Causal inference in medical records and complementary sys- tems pharmacology for metformin drug repurposing towards dementia. Nature Communications 13, 1 (2022), 7652

work page 2022

[5] [5]

Pengyu Cheng, Weituo Hao, Shuyang Dai, Jiachang Liu, Zhe Gan, and Lawrence Carin. 2020. Club: A contrastive log-ratio upper bound of mutual information. In International Conference on Machine Learning . PMLR, 1779–1788

2020

[6] [6]

Bénédicte Colnet, Imke Mayer, Guanhua Chen, Awa Dieng, Ruohong Li, Gaël Varoquaux, Jean-Philippe Vert, Julie Josse, and Shu Yang. 2024. Causal inference methods for combining randomized trials and observational studies: a review. Statist. Sci. 39, 1 (2024), 165–191

work page 2024

[7] [7]

Giovanni Corrao, Federica Nicotra, Andrea Parodi, Antonella Zambon, Davide Soranna, Franca Heiman, Luca Merlino, and Giuseppe Mancia. 2012. External adjustment for unmeasured confounders improved drug–outcome association estimates based on health care utilization data. Journal of Clinical Epidemiology 65, 11 (2012), 1190–1199

work page 2012

[8] [8]

Jacob Dorn, Kevin Guo, and Nathan Kallus. 2024. Doubly-valid/doubly-sharp sensitivity analysis for causal inference with unmeasured confounding. J. Amer. Statist. Assoc. (2024), 1–12

work page 2024