arxiv: 2604.08853 · v1 · submitted 2026-04-10 · 📊 stat.ME

Recognition: 3 theorem links

· Lean Theorem

The Illusion of Learning from Observational Data: An Empirical Bayes Perspective

Bohan Wu, David M. Blei, Donald P. Green, Sebastian Salazar

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:24 UTC · model grok-4.3

classification 📊 stat.ME

keywords empirical Bayesobservational studiescausal inferencecalibration studiesbias distributionshrinkage estimationillusion of learning

0 comments

The pith

Calibration studies identify the distribution of biases in observational data, allowing empirical Bayes shrinkage to recover causal effects consistently.

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

Observational studies often produce biased estimates of cause and effect because unmeasured confounders violate the assumptions needed for valid inference. The paper shows that this problem creates an illusion of learning unless prior information about the bias is available. Calibration studies, which target effects already known to be zero, make it possible to estimate the full distribution of such biases across many settings. Once that distribution is in hand, empirical Bayes methods can shrink the results from target observational studies toward the truth. With growing numbers of both calibration and observational studies, the bias distribution and the underlying causal parameters become consistently recoverable.

Core claim

Absent prior information about bias, observational results cannot meaningfully contribute to the estimation of a causal parameter. Calibration studies that target a causal effect known a priori to be zero identify the distribution of observational bias. This distribution then allows observational studies to inform the estimation of causal parameters via empirical Bayes shrinkage. With an increasing number of calibration and observation studies, both the bias distribution and the causal effect can be consistently recovered.

What carries the argument

Empirical Bayes shrinkage that uses a bias distribution estimated from calibration studies where the true causal effect is known to be zero.

If this is right

Observational studies can contribute to causal estimation once paired with calibration data that reveals the bias distribution.
Both the bias distribution and the causal effect become consistently recoverable as the number of calibration and observational studies increases.
Empirical Bayes shrinkage produces improved point estimates and uncertainty quantification for causal parameters from observational data.
The approach applies in settings where semi-synthetic data from real experiments can be used to validate performance.

Where Pith is reading between the lines

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

Fields with frequent null-effect experiments, such as certain policy evaluations, could systematically collect calibration data to make large observational archives more usable for causal questions.
The method suggests a practical design choice: embed null-effect probes within observational programs to anchor the bias distribution without additional randomized trials.
Extensions could examine how the number and design of calibration studies affect the rate at which causal estimates converge.

Load-bearing premise

The distribution of biases learned from calibration studies where the causal effect is known to be zero is the same as the bias distribution affecting the target observational studies of interest.

What would settle it

A simulation in which the true bias distribution in the target observational studies differs from the one recovered from calibration studies would cause the estimated causal effects to fail to converge to their true values as the number of studies grows.

Figures

Figures reproduced from arXiv: 2604.08853 by Bohan Wu, David M. Blei, Donald P. Green, Sebastian Salazar.

**Figure 4.** Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗

**Figure 5.** Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗

**Figure 5.** Figure 5 [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗

read the original abstract

Randomized experiments have long been the gold standard for scientists seeking to learn about cause and effect. When randomized experiments are infeasible, scientists often resort to observational studies, which are widely available and often large but rely on untestable assumptions that, when violated, may result in biased estimates. Uncertainty about bias leads to a phenomenon known as the illusion of learning from observational research (Gerber, Green and Kaplan, 2004a): absent prior information about bias, observational results cannot meaningfully contribute to the estimation of a causal parameter. To shatter the illusion, we take an empirical Bayes perspective. We show that the distribution of observational biases can be learned from calibration studies-experiments that target a causal effect that is known a priori to be zero. Calibration identifies the distribution of observational bias and allows observational studies to inform the estimation of causal parameters via empirical Bayes shrinkage. We formalize the illusion phenomenon in an empirical Bayes setting and show that, with an increasing number of calibration and observation studies, both the bias distribution and the causal effect can be consistently recovered. We illustrate our method through a simulation study and a semi-synthetic application based on Ferraro and Miranda (2013)'s water-usage experiment.

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.

Desk Editor's Note private letter to a colleague

The paper gives a clean empirical Bayes route to correct observational bias using separate calibration studies, with a consistency result, but the transportability of the learned bias distribution is a strong untested assumption.

read the letter

The core contribution is straightforward: they formalize the illusion of learning from observational data in an empirical Bayes framework, show that calibration studies (with known zero effects) identify the bias distribution, and prove that both the bias distribution and the target causal effect can be recovered consistently as the number of studies grows. They back this with simulations and a semi-synthetic example drawn from the Ferraro and Miranda water-usage experiment. That is new relative to the earlier Gerber-Green-Kaplan work on the illusion; the specific combination of calibration plus EB shrinkage plus the consistency theorem is not in the cited literature. The math looks internally consistent on its own terms, and the citation pattern is appropriate without obvious gaps or self-promotion. The illustrations are modest but on point for checking the method when assumptions hold. The main soft spot is exactly the one in the stress-test note. The whole procedure requires that the bias distribution estimated from calibration studies is the same one operating in the target observational studies. The paper treats this as given once the calibration studies are independent, but offers no robustness checks, sensitivity analysis, or discussion of what happens when selection, measurement, or context differ between the two sets. If that equality fails, the consistency result does not apply and the shrinkage can make things worse. That assumption is load-bearing, not minor. This is for causal-inference methodologists who already work with observational data in social science or policy settings and are looking for a principled way to bring in limited experimental calibration. It shows clear thinking and honest engagement with the problem. I would send it to peer review; the framework is novel enough and the theory is grounded enough that referees should see it, even though revisions will need to address the transportability issue directly.

Referee Report

2 major / 2 minor

Summary. The paper claims that the 'illusion of learning' from observational data—where untestable bias prevents meaningful contribution to causal parameter estimation—can be overcome via an empirical Bayes approach. Calibration studies (with known zero causal effects) are used to identify the distribution of observational biases, enabling shrinkage of observational estimates; with growing numbers of calibration and observational studies, both the bias distribution and causal effects are consistently recovered. The approach is formalized theoretically and illustrated with simulations plus a semi-synthetic application based on Ferraro and Miranda (2013).

Significance. If the core assumptions hold, the work offers a principled statistical framework for integrating abundant observational data with targeted calibration experiments to improve causal inference, directly addressing a known limitation in the literature. Strengths include the explicit EB formalization of the illusion phenomenon, the asymptotic consistency results, and the use of independent calibration data to ground the bias prior (avoiding direct circularity). The simulation and semi-synthetic checks provide initial empirical grounding.

major comments (2)

[§2–3] §2–3 (setup and assumptions): The central consistency claim (recovering both bias distribution and causal effect as the number of studies → ∞) rests on the assumption that the bias distribution learned from calibration studies (true effect = 0) is identical to the bias distribution operating in the target observational studies. This transportability assumption is stated without justification, sensitivity analysis, or discussion of plausible violations (e.g., differences in selection, measurement error, or context between calibration and target studies). If violated, the EB shrinkage step and the claimed recovery both fail even with infinite data.
[§4] §4 (consistency results): The theorems establish consistency under the maintained assumptions, but the paper does not examine finite-sample behavior, the rate of convergence, or robustness to mild misspecification of the bias distribution. Given that the practical value hinges on moderate numbers of studies, these omissions limit assessment of whether the method delivers reliable gains in realistic settings.

minor comments (2)

[Application section] The semi-synthetic application section would benefit from a clearer description of how the Ferraro and Miranda (2013) data were modified to create the observational and calibration components.
[Throughout] Notation for the bias distribution and the EB posterior could be introduced earlier and used consistently to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments, which help clarify the scope and limitations of our empirical Bayes framework. We address each major comment below and describe the revisions planned for the next version of the manuscript.

read point-by-point responses

Referee: [§2–3] The central consistency claim (recovering both bias distribution and causal effect as the number of studies → ∞) rests on the assumption that the bias distribution learned from calibration studies (true effect = 0) is identical to the bias distribution operating in the target observational studies. This transportability assumption is stated without justification, sensitivity analysis, or discussion of plausible violations (e.g., differences in selection, measurement error, or context between calibration and target studies). If violated, the EB shrinkage step and the claimed recovery both fail even with infinite data.

Authors: We agree that transportability of the bias distribution is a strong identifying assumption. The manuscript posits that calibration studies can be chosen to share the same bias-generating mechanisms as the target observational studies (e.g., comparable measurement protocols or selection processes). This mirrors standard transportability assumptions in causal inference and meta-analysis. We acknowledge that the current draft offers limited explicit justification and no sensitivity analysis. In the revision we will expand the assumptions section (§2–3) with additional discussion of when the assumption is plausible, add a subsection on potential violations (including differences in selection, measurement error, or context), and include a sensitivity analysis that perturbs the bias distribution to illustrate the consequences of violations. revision: yes
Referee: [§4] The theorems establish consistency under the maintained assumptions, but the paper does not examine finite-sample behavior, the rate of convergence, or robustness to mild misspecification of the bias distribution. Given that the practical value hinges on moderate numbers of studies, these omissions limit assessment of whether the method delivers reliable gains in realistic settings.

Authors: The consistency theorems in §4 are asymptotic results that hold as the number of studies diverges. The existing simulation study already reports performance for finite numbers of studies. To address the concern more directly, we will expand the simulation section with new experiments focused on moderate numbers of studies (e.g., 10–100), provide a brief discussion of the convergence rates implied by the theoretical analysis, and add robustness checks that introduce mild misspecification of the bias distribution to evaluate sensitivity in finite samples. revision: yes

Circularity Check

0 steps flagged

No significant circularity; calibration studies supply independent identification of the bias distribution.

full rationale

The derivation proceeds by positing a shared bias distribution across calibration studies (where the causal effect is known to be zero by design) and target observational studies, then applying standard empirical Bayes shrinkage and consistency arguments as the number of studies grows. The bias distribution is estimated from data external to the target observational studies, so the shrinkage step is not a fit to the quantities being corrected. No equation reduces a claimed prediction to a fitted input by construction, no uniqueness theorem is imported from the authors' own prior work, and the formalization of the illusion phenomenon builds on but does not tautologically rest upon the 2004 citation. The transportability assumption is stated explicitly rather than derived from the target data itself.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that calibration studies capture a transferable bias distribution; no new entities are introduced and no free parameters beyond the estimated bias distribution are required.

free parameters (1)

bias distribution
The distribution of observational biases is estimated empirically from the calibration studies and then used as the prior for shrinkage.

axioms (1)

domain assumption The bias distribution learned from zero-effect calibration studies applies to the observational studies whose causal effects are being estimated.
This transferability is required for the shrinkage step to correct the target estimates.

pith-pipeline@v0.9.0 · 5518 in / 1277 out tokens · 83059 ms · 2026-05-10T18:24:28.728145+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
We show that the distribution of observational biases can be learned from calibration studies—experiments that target a causal effect that is known a priori to be zero. ... fit μ̂,γ̂² by maximizing the marginal likelihood of the calibration studies
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
Theorem 4 (Calibration shatters the illusion). ... R(θ̂CEB,θ⋆)→0 as J,K→∞
IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean absolute_floor_iff_bare_distinguishability unclear
Assumption that the bias distribution identified from calibration studies ... is identical to the bias distribution operating in the target observational studies

Forward citations

Cited by 1 Pith paper

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

Empirical Bayes Rebiasing
stat.ME 2026-05 unverdicted novelty 6.0

Empirical Bayes rebiasing learns the bias distribution from paired noisy estimates to produce shorter calibrated intervals than full debiasing while maintaining coverage.

Reference graph

Works this paper leans on

55 extracted references · 14 canonical work pages · cited by 1 Pith paper

[1]

barticle [author] Abadie , A. A. Imbens , G. W. G. W. ( 2012 ). A martingale representation for matching estimators . Journal of the American Statistical Association 107 833-843 . 10.1080/01621459.2012.682537 barticle

work page doi:10.1080/01621459.2012.682537 2012
[2]

, Horvath , Hacsi T H

barticle [author] Anglemyer , Andrew A. , Horvath , Hacsi T H. T. Bero , Lisa L. ( 2014 ). Healthcare outcomes assessed with observational study designs compared with those assessed in randomized trials . Cochrane database of systematic reviews 4 . barticle

2014
[3]

bbook [author] Angrist , Joshua D J. D. Pischke , Jorn-Steffen J.-S. ( 2009 ). Mostly Harmless Econometrics: An Empiricist's Companion . Princeton University Press . bbook

2009
[4]

, Gerber , Alan S

barticle [author] Arceneaux , Kevin K. , Gerber , Alan S. A. S. Green , Donald P. D. P. ( 2010 ). A cautionary note on the use of matching to estimate causal effects: A n empirical example comparing matching estimates to an experimental benchmark . Sociological Methods and Research 39 . 10.1177/0049124110378098 barticle

work page doi:10.1177/0049124110378098 2010
[5]

Nickerson , David W

binproceedings [author] Arceneaux , Kevin K. Nickerson , David W. D. ( 2006 ). Put Your Money Where Your Mouth Is: A Field Experiment on Political Persuasion . In Annals of Voting and Elections Research . binproceedings

2006
[6]

, Imbens , Guido G

barticle [author] Athey , Susan S. , Imbens , Guido G. Wager , Stefan S. ( 2020 ). Approximate Residual Balancing: Debiased Inference of Average Treatment Effects in High Dimensions . Journal of the Royal Statistical Society Series B 82 597--607 . barticle

2020
[7]

Hartz , Arthur J

barticle [author] Benson , Kjell K. Hartz , Arthur J. A. J. ( 2000 ). A Comparison of Observational Studies and Randomized, Controlled Trials . New England Journal of Medicine 342 . 10.1056/nejm200006223422506 barticle

work page doi:10.1056/nejm200006223422506 2000
[8]

( 2013 )

bbook [author] Billingsley , Patrick P. ( 2013 ). Convergence of Probability Measures . John Wiley & Sons . bbook

2013
[9]

bbook [author] Bloom , Howard S H. S. ( 2005 ). Learning More from Social Experiments: Evolving Analytic Approaches . Russell Sage Foundation . bbook

2005
[10]

bbook [author] Brady , Henry E H. E. Collier , David D. ( 2010 ). Rethinking Social Inquiry: Diverse Tools, Shared Standards . Bloomsbury Publishing PLC . bbook

2010
[11]

Hunter , Albert A

bbook [author] Brewer , John J. Hunter , Albert A. ( 2006 ). Foundations of Multimethod Research: Synthesizing Styles . Sage . bbook

2006
[12]

barticle [author] Cole , Stephen R. S. Stuart , Elizabeth A. E. ( 2010 ). Generalizing Evidence from Randomized Trials to Target Populations: The Actg 320 Trial . American Journal of Epidemiology 172 107--115 . barticle

2010
[13]

, Shah , Nirav N

bincollection [author] Concato , John J. , Shah , Nirav N. Horwitz , Ralph I R. I. ( 2017 ). Randomized, controlled trials, observational studies, and the hierarchy of research designs . In Research Ethics 207--212 . Routledge . bincollection

2017
[14]

bbook [author] Creswell , John W J. W. Creswell , J David J. D. ( 2017 ). Research Design: Qualitative, Quantitative, and Mixed Methods Approaches . Sage publications . bbook

2017
[15]

barticle [author] Curtis , Jeffrey R J. R. , Larson , Joseph C J. C. , Delzell , Elizabeth E. , Brookhart , Maurice Alan M. A. , Cadarette , Suzanne M S. M. , Chlebowski , Rowan R. , Judd , Suzanne S. , Safford , Monika M. , Solomon , Daniel H D. H. LaCroix , Andrea Z A. Z. ( 2011 ). Placebo adherence, clinical outcomes, and mortality in the women's healt...

2011
[16]

barticle [author] Dehejia , Rajeev H. R. H. Wahba , Sadek S. ( 1999 ). Causal Effects in Non-Experimental Studies: Reevaluating the Evaluation of Training Programs . Journal of the American Statistical Association 94 1053--1062 . barticle

1999
[17]

( 2024 )

bbook [author] Ding , Peng P. ( 2024 ). A First Course in Causal Inference . Chapman and Hall/CRC . bbook

2024
[18]

( 2012 )

bbook [author] Efron , Bradley B. ( 2012 ). Large-scale Inference: Empirical Bayes Methods for Estimation, Testing, and Prediction 1 . Cambridge University Press . bbook

2012
[19]

( 2014 )

barticle [author] Efron , Bradley B. ( 2014 ). Two modeling strategies for empirical B ayes estimation . Statistical Science 29 . 10.1214/13-STS455 barticle

work page doi:10.1214/13-sts455 2014
[20]

barticle [author] Ferraro , Paul J. P. J. Miranda , Juan Jos \'e J. J. ( 2013 ). Heterogeneous treatment effects and mechanisms in information-based environmental policies: Evidence from a large-scale field experiment . Resource and Energy Economics 35 356--379 . barticle

2013
[21]

Hill , Jennifer J

bbook [author] Gelman , Andrew A. Hill , Jennifer J. ( 2006 ). Data Analysis Using Regression and Multilevel/Hierarchical Models . Cambridge University Press . bbook

2006
[23]

binbook [author] Gerber , Alan S. A. S. , Green , Donald P. D. P. Kaplan , Edward H. E. H. ( 2004 b). The illusion of learning from observational research . 10.1017/CBO9780511492174.012 binbook

work page doi:10.1017/cbo9780511492174.012 2004
[24]

barticle [author] Gerber , Alan S A. S. , Green , Donald P D. P. , Kaplan , Edward H E. H. Kern , Holger L H. L. ( 2010 ). Baseline, placebo, and treatment: E fficient estimation for three-group experiments . Political Analysis 18 297--315 . barticle

2010
[25]

barticle [author] Gordon , Brett R B. R. , Zettelmeyer , Florian F. , Bhargava , Neha N. Chapsky , Dan D. ( 2019 ). A comparison of approaches to advertising measurement: E vidence from big field experiments at F acebook . Marketing Science 38 193--225 . barticle

2019
[26]

barticle [author] Guo , Xinyu X. et al. ( 2021 a). Control Variate Methods for Combining Experimental and Observational Data in Causal Inference . Journal of Causal Inference 9 123--145 . barticle

2021
[27]

, Wang , Serena S

barticle [author] Guo , Wenshuo W. , Wang , Serena S. , Ding , Peng P. , Wang , Yixin Y. Jordan , Michael I M. I. ( 2021 b). Multi-source causal inference using control variates . arXiv preprint arXiv:2103.16689 . barticle

work page arXiv 2021
[28]

barticle [author] Hirano , K. K. , Imbens , G. W. G. W. Ridder , G. G. ( 2003 ). Efficient estimation of average treatment effects using the estimated propensity score . Econometrica 71 1161-1189 . 10.1111/1468-0262.00442 barticle

work page doi:10.1111/1468-0262.00442 2003
[29]

bbook [author] Ignatiadis , N. N. Sen , B. B. ( 2025 ). Empirical B ayes: From H erbert R obbins to Modern Theory and Applications . bbook

2025
[30]

bbook [author] Imbens , Guido W G. W. Rubin , Donald B D. B. ( 2015 ). Causal Inference in Statistics, Social, and Biomedical Sciences . Cambridge University Press . bbook

2015
[31]

, Kallus , Nathan N

barticle [author] Imbens , Guido G. , Kallus , Nathan N. , Mao , Xiaojie X. Wang , Yuhao Y. ( 2025 ). Long-term causal inference under persistent confounding via data combination . Journal of the Royal Statistical Society Series B: Statistical Methodology 87 362--388 . barticle

2025
[32]

barticle [author] Jiang , W. W. Zhang , C. H. C. H. ( 2009 ). General maximum likelihood empirical B ayes estimation of normal means . Annals of Statistics 37 1647-1684 . 10.1214/08-AOS638 barticle

work page doi:10.1214/08-aos638 2009
[33]

barticle [author] Kiefer , J. J. Wolfowitz , J. J. ( 1956 ). Consistency of the Maximum Likelihood Estimator in the Presence of Infinitely Many Incidental Parameters . The Annals of Mathematical Statistics 27 887-906 . 10.1214/AOMS/1177728066 barticle

work page doi:10.1214/aoms/1177728066 1956
[34]

barticle [author] LaLonde , Robert J. R. ( 1986 ). Evaluating the Econometric Evaluations of Training Programs with Experimental Data . American Economic Review 76 604--620 . barticle

1986
[35]

J., and Ding, P

barticle [author] Lin , Zhexiao Z. , Bickel , Peter J P. J. Ding , Peng P. ( 2026 ). Introducing the b-value: combining unbiased and biased estimators from a sensitivity analysis perspective . arXiv preprint arXiv:2602.16310 . barticle

work page arXiv 2026
[36]

, Tchetgen , Eric Tchetgen E

barticle [author] Lipsitch , Marc M. , Tchetgen , Eric Tchetgen E. T. Cohen , Ted T. ( 2010 ). Negative controls: a tool for detecting confounding and bias in observational studies . Epidemiology 21 383-388 . barticle

2010
[37]

, Tchetgen Tchetgen , Eric J

barticle [author] Miao , Xian X. , Tchetgen Tchetgen , Eric J. E. J. et al. ( 2018 ). Identification of Causal Effects with Auxiliary Variables: The Role of Negative Controls . Journal of the American Statistical Association 113 172--186 . barticle

2018
[38]

( 1923 )

barticle [author] Neyman , Jerzy J. ( 1923 ). On the application of probability theory to agricultural experiments. E ssay on principles . Ann. Agricultural Sciences 1--51 . barticle

1923
[39]

( 1956 )

binproceedings [author] Robbins , Herbert H. ( 1956 ). An empirical B ayes approach to statistics . In Proceedings of the 3rd Berkeley Symposium on Mathematical Statistics and Probability 157--163 . binproceedings

1956
[40]

van der Laan , M J M

barticle [author] Rose , S S. van der Laan , M J M. J. ( 2011 ). Targeted Learning: Causal Inference for Observational and Experimental Data . Targeted Learning: Causal Inference for Observational and Experimental Data . barticle

2011
[41]

barticle [author] Rosenbaum , Paul R P. R. Rubin , Donald B D. B. ( 1983 ). The central role of the propensity score in observational studies for causal effects . Biometrika 70 41-55 . barticle

1983
[42]

barticle [author] Rosenman , Evan TR E. T. , Basse , Guillaume G. , Owen , Art B A. B. Baiocchi , Mike M. ( 2023 ). Combining observational and experimental datasets using shrinkage estimators . Biometrics 79 2961--2973 . barticle

2023
[43]

barticle [author] Smith , J. A. J. A. Todd , P. E. P. E. ( 2001 ). Reconciling conflicting evidence on the performance of propensity-score matching methods . American Economic Review 91 112-118 . 10.1257/aer.91.2.112 barticle

work page doi:10.1257/aer.91.2.112 2001
[44]

barticle [author] Soloff , Jake A J. A. , Guntuboyina , Adityanand A. Sen , Bodhisattva B. ( 2025 ). Multivariate, heteroscedastic empirical B ayes via nonparametric maximum likelihood . Journal of the Royal Statistical Society Series B: Statistical Methodology 87 1--32 . barticle

2025
[45]

barticle [author] Stuart , Elizabeth A. E. ( 2011 ). Matching Methods for Causal Inference: A Review and a Look Forward . Statistical Science 25 1--21 . barticle

2011
[46]

barticle [author] Tchetgen Tchetgen , Eric J. E. J. ( 2014 ). Using Negative Controls to Adjust for Unmeasured Confounding . Epidemiology 25 364--380 . barticle

2014
[47]

Jackson , Dave D

barticle [author] Turner , Ruth R. Jackson , Dave D. ( 2009 ). A Hierarchical B ayesian Approach to Bias Adjustment in Meta-Analysis . Statistics in Medicine 28 331--347 . barticle

2009
[48]

bbook [author] Van der Vaart , Aad W A. W. ( 2000 ). Asymptotic Statistics 3 . Cambridge university press . bbook

2000
[49]

barticle [author] van der Vaart , A. W. A. W. Wellner , Jon A. J. A. ( 2023 ). Weak Convergence and Empirical Processes . 10.1007/978-3-031-29040-4 barticle

work page doi:10.1007/978-3-031-29040-4 2023
[50]

barticle [author] Waddington , Hugh Sharma H. S. , Villar , Paul Fenton P. F. Valentine , Jeffrey C J. C. ( 2023 ). Can non-randomised studies of interventions provide unbiased effect estimates? A systematic review of internal replication studies . Evaluation Review 47 563--593 . barticle

2023
[51]

barticle [author] Wong , Vivian C V. C. , Steiner , Peter M P. M. Anglin , Kylie L K. L. ( 2018 ). What can be learned from empirical evaluations of nonexperimental methods? Evaluation review 42 147--175 . barticle

2018
[52]

barticle [author] Xie , X. X. , Kou , S. C. S. C. Brown , L. D. L. D. ( 2012 ). SURE estimates for a heteroscedastic hierarchical model . Journal of the American Statistical Association 107 1465-1479 . 10.1080/01621459.2012.728154 barticle

work page doi:10.1080/01621459.2012.728154 2012
[53]

, Koenecke , Allison A

barticle [author] Xiong , Ruoxuan R. , Koenecke , Allison A. , Powell , Michael M. , Shen , Zhu Z. , Vogelstein , Joshua T J. T. Athey , Susan S. ( 2023 ). Federated causal inference in heterogeneous observational data . Statistics in Medicine 42 4418--4439 . barticle

2023
[54]

Ding , Peng P

barticle [author] Yang , Shu S. Ding , Peng P. ( 2020 ). Combining multiple observational data sources to estimate causal effects . Journal of the American Statistical Association . barticle

2020
[55]

, Gao , Chenyin C

barticle [author] Yang , Shu S. , Gao , Chenyin C. , Zeng , Donglin D. Wang , Xiaofei X. ( 2023 ). Elastic integrative analysis of randomised trial and real-world data for treatment heterogeneity estimation . Journal of the Royal Statistical Society Series B: Statistical Methodology 85 575--596 . barticle

2023
[56]

I., and Imbens, G

barticle [author] Yang , Xuelin X. , Lin , Licong L. , Athey , Susan S. , Jordan , Michael I M. I. Imbens , Guido W G. W. ( 2025 ). Cross-Validated Causal Inference: a Modern Method to Combine Experimental and Observational Data . arXiv preprint arXiv:2511.00727 . barticle

work page arXiv 2025