Pith. sign in

REVIEW 3 major objections 4 minor 49 references

An adaptive design for optimizing treatment assignment in randomized clinical trials

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a two-stage adaptive design, which re-optimizes the treatment allocation ratio at an interim analysis using estimated outcome-variance functions, yields a treatment-effect estimator asymptotically equivalent to an orac

desk verdict Solid adaptive-design methods paper with real novelty; the main efficiency guarantee is heuristic rather than proven, but the paper is honest about it and the simulations back it up. read the letter →

arxiv 2509.00429 v1 pith:UGVNEMGX submitted 2025-08-30 stat.ME

Wei Zhang , Zhiwei Zhang , Aiyi Liu This is my paper
classification stat.ME MSC 62L0562F1262K05
keywords adaptivedesigntreatmentallocationconditionalvariancefunctionscovariateadjustmentaugmentedestimatorspropensityscoreclinicaltrialefficiencyasymptoticnormality
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

The paper proposes a practical answer to a design dilemma: the most efficient treatment-allocation rule in a randomized trial depends on outcome variances conditional on baseline covariates, which are usually unknown before the trial starts. The adaptive design starts with simple or cautiously optimized randomization, re-estimates those variance functions at an interim analysis, and switches to an optimized allocation for the remaining patients. Because the adaptation makes data dependent across stages and non-identically distributed, the paper develops a class of augmented treatment-effect estimators with explicit optimal weights and augmentation functions. The main result is that substituting estimated weights and functions leaves the estimator asymptotically normal and equivalent to the oracle estimator, so the adaptive design captures most of the efficiency that would require knowing the variance functions in advance. Simulations and a stroke-trial reanalysis indicate the efficiency gain can be substantial when prior information is scarce.

What carries the argument

The central objects are augmented estimators δ̂_aug(b1, b2, θ), which combine stage-specific treatment-arm averages with covariate correction terms (A − π)b(W), and their AIPW analogues under covariate-dependent randomization. The optimal augmentation functions solve a variance-minimization problem and equal weighted contrasts of the conditional mean functions m_a(W) = E[Y(a)|W]; the optimal stage-combination weight θ*_opt is an inverse-variance weighted average of stage-specific variances. Plugging in estimates of these quantities makes the estimator first-order equivalent to the fixed-limit oracle estimator.

What would settle it

Run the two-stage design with a working variance model whose fitted π2 converges in probability to a value farther from the true optimal π_opt than π1 is; if the oracle-equivalence claim is right, the actual variance of δ̂_aug should still match the limiting formula, but if the adaptive design then underperforms the one-stage design, the claimed efficiency guarantee fails. A sharper test: simulate a case where π2 does not converge (oscillating estimates) and check whether nominal 95% intervals maintain coverage.

Watch

Extended reading notes

Core claim

The paper's central claim is that the proposed two-stage adaptive estimator—denoted δ̂_aug with estimated augmentation functions and estimated optimal weight—is consistent and asymptotically normal, and is asymptotically equivalent to the oracle estimator built on the limiting augmentation functions and the limiting optimal weight (Theorem 2). When the fitted outcome-regression limit equals the true conditional mean, this estimator attains the smallest asymptotic variance among all estimators in the considered class. The same conclusion holds for the AIPW analogue under covariate-dependent randomization and for the multi-stage extension.

Load-bearing premise

The stage-1 estimates of the conditional variance functions must converge so that the resulting stage-2 allocation probability π2 stabilizes at a limit π*2 that is no less efficient than the initial allocation; the paper expects this to hold but does not give primitive conditions, and Appendix D shows misspecified variance models can erode the gain.

Editorial extensions

If this is right

  • A trial with no prior variance information can start with 1:1 randomization and still approach the precision of a trial that knew the optimal allocation from the start.
  • Estimated optimal weights and augmentation functions yield valid asymptotic inference, so reported standard errors and confidence intervals are reliable under the stated conditions.
  • The same estimation strategy extends to covariate-dependent randomization and to more than two stages, making hybrid designs such as optimized CIR followed by optimized CDR practical.
  • The efficiency gain occurs when the second-stage allocation limit is closer to the true optimal allocation than the first-stage allocation; otherwise the adaptive design offers little or no benefit.
  • Simulations with binary outcomes and logistic working models show roughly 10–20% relative efficiency gains and near-nominal confidence-interval coverage.

Reading between the lines

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

  • The convergence-in-probability condition on the updated allocation probability suggests a general template: any design adaptation whose tuning parameters stabilize quickly could be handled with the same oracle-equivalence proof, for instance sample-size re-estimation.
  • The simulation evidence that CIR adaptation is more robust than CDR under severe working-model misspecification suggests practitioners should prefer CIR when variance estimates are questionable and reserve CDR for settings with adequate interim data.
  • The explicit optimal stage weight θ*_opt gives a principled way to choose how many patients to enroll before the interim analysis: the first stage should be large enough to make the second-stage variance smaller than the first-stage variance.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper develops an adaptive two-stage (and multi-stage) randomized trial design in which the treatment allocation mechanism is updated at an interim analysis using estimates of conditional variance functions of the potential outcomes. For both covariate-independent randomization (CIR) and covariate-dependent randomization (CDR), the authors define a class of weighted, augmented treatment-effect estimators, derive the asymptotic variance under a limiting allocation probability, and characterize the optimal augmentation functions and stage-combination weight. They show that plugging in estimated augmentation functions and weights preserves consistency and asymptotic normality, and that the resulting estimator is asymptotically equivalent to the oracle estimator in the class. The methods are evaluated by simulation and illustrated on the NINDS rt-PA trial. Theorems 1 and 2 and the CDR/multi-stage extensions are proved in Appendix C of the supplementary materials.

Significance. If the claims hold, the paper provides a useful and principled framework for incorporating accruing information about outcome-variance structure into the treatment-allocation design of a randomized trial, going beyond static optimal designs. The main theoretical contributions—the asymptotic distribution under a data-dependent second-stage allocation, the explicit optimal augmentation functions, and the consistency of the estimated variance estimator—are nontrivial and appear broadly correct. The simulation study is reasonably extensive, including misspecification, smaller samples, and different stage allocations, and the real-data illustration is appropriate. The paper is honest in labeling the cross-stage efficiency gain as heuristic (Remark 1) and in reporting settings where gains are eroded. However, a key proof step in Theorem 2 is incomplete as written, and the central efficiency advantage over one-stage designs rests on informal conditions rather than a theorem. These issues are fixable but require attention.

major comments (3)
  1. [Appendix C, Proof of Theorem 2] The Chebyshev step used to show that the augmentation difference is asymptotically negligible is incorrect as stated. The display equating E[(n^{-1/2} Σ_i (A_i−π)(êb_1(W_i)−b_1(W_i)))^2] with E[(A_i−π)^2(êb_1(W)−b_1(W))^2] ignores the dependence among summands induced by êb_1 being estimated from the same data. L2 convergence of êb_1 to b_1 does not by itself eliminate the cross terms in the variance of the sample mean. This step is load-bearing for the asymptotic equivalence in Theorem 2. Please supply a valid argument (e.g., Donsker/Glivenko–Cantelli conditions on the fitted function class, or sample-splitting), or state additional conditions under which the displayed equality holds.
  2. [Section 4, Remark 1] The paper's central practical motivation—that the adaptive design improves efficiency over a one-stage design—is not established. Remark 1 gives the criterion σ*2_2,cir < σ*2_1,cir, but no primitive conditions ensure that the limiting second-stage allocation π*2 is closer to π_opt than π1 is. Because π2 is obtained by plugging estimated variance functions into the optimal allocation formula, a misspecified working model can yield a π*2 that is no better than, or worse than, π1. Appendix D's severe-misspecification results (Table S5) show erosion of gains but do not identify a uniformly guaranteed improvement. Please either provide sufficient conditions for the key inequality, or explicitly state that the efficiency gain is heuristic and supported only by simulations, and discuss the risk of efficiency loss under misspecification.
  3. [Section 4 / Appendix C] The regularity conditions in Theorem 1 are informal: convergence of π2 to π*2 is assumed in probability, and the text states this is 'expected to hold' if the underlying estimates converge, without primitive conditions. Similarly, the convergence of (bm1, bm0) to (m*1, m*0) in Appendix C is described only as holding 'under certain regularity conditions.' Since the plug-in asymptotics and the variance estimator consistency both rely on these assumptions, the authors should state explicit, verifiable sufficient conditions for the main theorems, at least for the working models used in the simulations (e.g., M-estimators for generalized linear models with bounded moments and Glivenko–Cantelli conditions).
minor comments (4)
  1. [Section 5 / Table 3] In the application, relative efficiencies are computed from estimated variances for a single generated dataset, not from repeated sampling. The text should make this clearer so readers do not interpret the relative efficiency values as empirical sampling results.
  2. [Appendix D, Table S5] The severe-misspecification simulation omits the treatment-by-covariate interaction from the working model. It would be useful to report the actual limiting π*2 values (or their estimates) for this scenario, to directly illustrate when the inequality in Remark 1 fails.
  3. [References] There is a typo in the van der Laan and Robins reference: 'Spring-Verlag' should be 'Springer-Verlag'.
  4. [Section 3] The sentence 'Whether π2 is optimized for all of W or a coarsened version of it has no impact on the subsequent development' could be misread as implying that the choice of X does not affect efficiency. The subsequent development is indeed unaffected, but the efficiency of the design does depend on the choice of X; a brief clarification would help.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the adaptive-design estimation theory is derived from first principles; self-citations to Zhang et al. (2023) are external support, not fitted inputs.

full rationale

The paper's central derivation chain is self-contained. Theorem 1 is proved in Appendix C by expanding the estimator, applying the CLT to stage-1 sums and, conditional on D1, to stage-2 sums, and then passing through the assumed limit π2 → π*2. The optimal augmentation functions (b1,opt, b2,opt) are obtained by directly minimizing the displayed variance expression var{ψ1^aug} + λ var{ψ2^aug}; the optimal weight θ*opt is obtained by minimizing θ^2σ*2_1,cir + λ(1−θ)^2σ*2_2,cir. Theorem 2 then follows from Slutsky/Chebyshev arguments for plug-in estimators. No fitted parameter is renamed as a prediction: the estimated allocation π2 is not used as evidence for the estimator's optimality, and the consistency/normality results do not require π2 to equal the true optimum. The only reliance on the authors' earlier work (Zhang et al., 2023) is for the single-stage optimal-design formulas (1)–(2) and the global-minimum statement in Remark 1; these are external, previously published results with stated assumptions, not consequences of the present adaptive procedure, and they are not fitted to the present simulations or data. Remark 1 explicitly labels the efficiency gain 'Heuristically' and conditions it on σ*2_2,cir < σ*2_1,cir, so the paper does not present the gain as a theorem forced by its own definitions. The simulations and the NINDS illustration are empirical demonstrations, not inputs to the estimator construction. Overall, no circular step in the sense of Eq. X = Eq. Y by construction, fitted input called prediction, or a load-bearing self-citation chain was found.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The theoretical framework adds no new entities or ad hoc parameters. The free parameters of the simulation study (e.g., logistic regression coefficients in the working model) are nuisance estimation targets, not free parameters of the proposed methodology. The key additional assumptions are convergence and correct-specification conditions needed for the asymptotic optimality claims.

assumptions (5)
  • standard math Potential outcomes and randomization: Y = A Y(1) + (1-A) Y(0), and A is independent of (Y(1), Y(0)) given W under CIR or CDR.
    Invoked in Section 2 to define the estimand and the treatment assignment mechanism.
  • standard math Finite second moments E{Y(a)^2} < ∞ and smooth link function g with derivative g'.
    Used in Section 2 and in the Taylor expansions and central limit arguments in the proof of Theorem 1 (Appendix C).
  • domain assumption π2 converges in probability to π*2 ∈ (0,1) and n1/n2 converges to λ ∈ (0,∞) as n1→∞.
    Imposed in Theorem 1 (Section 4) and needed for the asymptotic variance formula; the paper states this is expected if variance estimates converge.
  • domain assumption Estimators (bµ1,bµ0) and (bm1,bm0) converge in probability to (µ1,µ0) and limits (m*1,m*0).
    Used in Section 4 and Appendix C to justify plug-in optimality of the augmentation functions and consistency of bθ.
  • domain assumption For the global efficiency claim, the outcome regression limit equals the truth: (m*1,m*0) = (m1,m0).
    Needed in Theorem 2 and Remark 1 to assert the estimator attains the smallest variance among all estimators in the class; in misspecified models the estimator remains consistent but may not be optimal.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An adaptive design for optimizing treatment assignment in randomized clinical trials." pith.science (2026). https://pith.science/paper/UGVNEMGX

@misc{pith2026250900429,
  author       = {Pith},
  title        = {Pith review of: An adaptive design for optimizing treatment assignment in randomized clinical trials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UGVNEMGX}},
  note         = {Machine review of arXiv:2509.00429}
}
read the original abstract

The treatment assignment mechanism in a randomized clinical trial can be optimized for statistical efficiency within a specified class of randomization mechanisms. Optimal designs of this type have been characterized in terms of the variances of potential outcomes conditional on baseline covariates. Approximating these optimal designs requires information about the conditional variance functions, which is often unavailable or unreliable at the design stage. As a practical solution to this dilemma, we propose a multi-stage adaptive design that allows the treatment assignment mechanism to be modified at interim analyses based on accruing information about the conditional variance functions. This adaptation has profound implications on the distribution of trial data, which need to be accounted for in treatment effect estimation. We consider a class of treatment effect estimators that are consistent and asymptotically normal, identify the most efficient estimator within this class, and approximate the most efficient estimator by substituting estimates of unknown quantities. Simulation results indicate that, when there is little or no prior information available, the proposed design can bring substantial efficiency gains over conventional one-stage designs based on the same prior information. The methodology is illustrated with real data from a completed trial in stroke.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 47 canonical work pages

  1. [1]

    S., Shao, J., Liu, J., Du, Y., Yi, Y

    Bannick, M. S., Shao, J., Liu, J., Du, Y., Yi, Y. and Ye, T. (2025) A general form of covariate adjustment in clinical trials under covariate-adaptive randomization. Biometrika, 112, asaf029

    work page 2025

  2. [2]

    (1993) Efficient and Adaptive Estimation for Semiparametric Models

    Bickel, P.J., Klaassen, C.A.J., Ritov, Y and Wellner, J.A. (1993) Efficient and Adaptive Estimation for Semiparametric Models. Baltimore, MD: Johns Hopkins University Press

    work page 1993

  3. [3]

    and et al

    Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and et al. (2018) Double/debiased machine learning for treatment and structural parameters. Econometrics Journal, 21, C1--C68

    work page 2018

  4. [4]

    and Carroll, R

    Davidian, M. and Carroll, R. J. (1987) Variance function estimation. Journal of the American Statistical Association, 82, 1079--1091

    work page 1987

  5. [5]

    and Nyquist, H

    Fackle-Fornius, E. and Nyquist, H. (2015) Optimal allocation to treatment groups under variance heterogeneity. Statistica Sinica, 25, 537--549

    work page 2015

  6. [6]

    Guidance for Industry: Adjusting for covariates in randomized clinical trials for drugs and biological products

    Food and Drug Administration (2023). Guidance for Industry: Adjusting for covariates in randomized clinical trials for drugs and biological products. Available at !https://www.fda.gov/regulatory-information/search-fda-guidance-documents/! !adjusting-covariates-randomized-clinical-trials-drugs-and-biological-products!

    work page 2023

  7. [7]

    and Friedman, J

    Hastie, T., Tibshirani, R. and Friedman, J. (2009) The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd ed. New York, Springer-Verlag

    work page 2009

  8. [8]

    and Karlan, D

    Hahn, J., Hirano, K. and Karlan, D. (2011) Adaptive experimental design using the propensity score. Journal of Business & Economic Statistics, 29, 96--108

    work page 2011

Show all 49 references
  1. [9]

    Ingall, T.J., O’Fallon, W.M., Asplund, K., Goldfrank, L.R., Hertzberg, V.S., Louis, T.A. et al. (2004) Findings from the reanalysis of the NINDS tissue plasminogen activator for acute ischemic stroke treatment trial. Stroke, 35, 2418--2424

    2004

  2. [10]

    and G'Sell, M

    Kennedy, E.H., Balakrishnan, S. and G'Sell, M. (2020) Sharp instruments for classifying compliers and generalizing causal effects. Annals of Statistics, 48, 2008--2030

    2020

  3. [11]

    and van der Laan, M.J

    Moore, K.L. and van der Laan, M.J. (2009) Covariate adjustment in randomized trials with binary outcomes: targeted maximum likelihood estimation. Statistics in Medicine, 28, 39--64

    2009

  4. [12]

    Neyman, J. (1934). On the two different aspects of the representative method: The method of stratified sampling and the method of purposive selection. Journal of the Royal Statistical Society, 97, 558--625

    1934

  5. [13]

    (1995) Tissue plasminogen activator for acute ischemic stroke

    NINDS rt-PA Stroke Study Group. (1995) Tissue plasminogen activator for acute ischemic stroke. New England Journal of Medicine, 333, 1581--1587

    1995

  6. [14]

    and van der Laan, M.J

    Polley, E.C., Rose, S. and van der Laan, M.J. (2011) Super learning. In Targeted Learning, pages 43--66. New York, Springer

    2011

  7. [15]

    (1983) Clinical trials: a practical approach

    Pocock, S.J. (1983) Clinical trials: a practical approach. New York, John Wiley and Sons

    1983

  8. [16]

    and Brumback, B

    Robins, J.M., Hernan, M.A. and Brumback, B. (2000) Marginal Structural models and causal inference in epidemiology. Epidemiology, 11, 550--560

    2000

  9. [17]

    and Rubin, D.B

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

    1983

  10. [18]

    and Rubin, D.B

    Rosenbaum, P.R. and Rubin, D.B. (1984) Reducing bias in observational studies using subclassification on the propensity score. Journal of the American Statistical Association, 79, 516--524

    1984

  11. [19]

    and Rubin, D.B

    Rosenbaum, P.R. and Rubin, D.B. (1985) Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. The American Statistician, 39, 33--38

    1985

  12. [20]

    and Sverdlov, O

    Rosenberger, W.F. and Sverdlov, O. (2008) Handling covariates in the design of clinical trials. Statistical Science, 23, 404--419

    2008

  13. [21]

    and Simon, R

    Pocock, S.J. and Simon, R. (1975) Sequential treatment assignment with balancing for prognostic factors in the controlled clinical trial. Biometrics, 31, 103--115

    1975

  14. [22]

    and van der Laan, M.J

    Rosenblum, M. and van der Laan, M.J. (2010) Simple, efficient estimators of treatment effects in randomized trials using generalized linear models to leverage baseline variables. International Journal of Biostatistics, 6, article 13

    2010

  15. [23]

    (2023) Efficient semiparametric estimation of average treatment effects under covariate-adaptive randomization

    Rafi, A. (2023) Efficient semiparametric estimation of average treatment effects under covariate-adaptive randomization. Available at https://arxiv.org/pdf/2305.08340

    2023 arXiv

  16. [24]

    (1974) Estimating causal effects of treatments in randomized and nonrandomized studies

    Rubin, D.B. (1974) Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology, 66, 688--701

    1974

  17. [25]

    and van der Laan, M.J

    Rubin, D.B. and van der Laan, M.J. (2008) Empirical efficiency maximization: improved locally efficient covariate adjustment in randomized experiments and survival analysis. International Journal of Biostatistics, 4, article 5

    2008

  18. [26]

    and Wei, L.J

    Tian, L., Cai, T., Zhao, L. and Wei, L.J. (2012) On the covariate-adjusted estimation for an overall treatment difference with data from a randomized comparative clinical trial. Biostatistics, 13, 256--273

    2012

  19. [27]

    (2006) Semiparametric Theory and Missing Data

    Tsiatis, A.A. (2006) Semiparametric Theory and Missing Data. New York, Springer

    2006

  20. [28]

    and Lu, X

    Tsiatis, A.A., Davidian, M., Zhang, M. and Lu, X. (2008) Covariate adjustment for two-sample treatment comparisons in randomized clinical trials: a principled yet flexible approach. Statistics in Medicine, 27, 4658--4677

    2008

  21. [29]

    and Hubbard, A.E

    van der Laan, M.J., Polley, E.C. and Hubbard, A.E. (2007) Super Learner. Statistical Applications in Genetics and Molecular Biology, 6, article 5

    2007

  22. [30]

    and Robins, J.M

    van der Laan, M.J. and Robins, J.M. (2003) Unified Methods for Censored Longitudinal Data and Causality. New York, Spring-Verlag

    2003

  23. [31]

    and Rose, S

    van der Laan, M.J. and Rose, S. (2011) Targeted Learning: Causal Inference for Observational and Experimental Data. New York, Springer

    2011

  24. [32]

    (1998) Asymptotic Statistics

    van der Vaart, A.W. (1998) Asymptotic Statistics. Cambridge, Cambridge University Press

    1998

  25. [33]

    and Wellner, J.A

    van der Vaart, A.W. and Wellner, J.A. (1996) Weak Convergence and Empirical Processes with Applications to Statistics. New York, Springer-Verlag

    1996

  26. [34]

    and Zhu, W

    Wong, W.K. and Zhu, W. (2008) Optimum treatment allocation rules under a variance heterogeneity model. Statistics in Medicine, 27, 4581--4595

    2008

  27. [35]

    and Zhang, H

    Zhang, Z., Li, W. and Zhang, H. (2020) Efficient estimation of Mann-Whitney-type effect measures for right-censored survival outcomes in randomized clinical trials. Statistics in Biosciences, 12, 246--262

    2020

  28. [36]

    (2019) Machine learning methods for leveraging baseline covariate information to improve the efficiency of clinical trials

    Zhang, Z and Ma, S. (2019) Machine learning methods for leveraging baseline covariate information to improve the efficiency of clinical trials. Statistics in Medicine, 38, 1703--1714

    2019

  29. [37]

    and Davidian, M

    Zhang, M., Tsiatis, A.A. and Davidian, M. (2008) Improving efficiency of inferences in randomized clinical trials using auxiliary covariates. Biometrics, 64, 707--715

    2008

  30. [38]

    and Liu, A

    Zhang, W., Zhang, Z. and Liu, A. (2023) Optimizing treatment allocation in randomized clinical trials by leveraging baseline covariates. Biometrics, 79, 2815--2829

    2023

  31. [39]

    (2025) A connection between covariate adjustment and stratified randomization in randomized clinical trials

    Zhang, Z. (2025) A connection between covariate adjustment and stratified randomization in randomized clinical trials. Statistical Methods in Medical Research, 34, 2829--844

    2025

  32. [40]

    (2023) Adaptive Neyman Allocation

    Zhao, J. (2023) Adaptive Neyman Allocation. arXiv preprint arxiv:2309.08808

    2023

  33. [41]

    and van der Laan, M.J

    Zheng, W. and van der Laan, M.J. (2011) Cross-validated targeted minimum-loss-based estimation. In Targeted Learning, pages 459--474. New York, Springer

    2011

  34. [42]

    and Rosenblum, M

    Wang, B., Ogburn, E.L. and Rosenblum, M. (2019) Analysis of covariance in randomized trials: more precision and valid confidence intervals, without model assumptions. Biometrics, 75, 1391--1400

    2019

  35. [43]

    Amin-Esmaeili, M

    Wang, B., Susukida, R., Mojtabai, R. Amin-Esmaeili, M. and Rosenblum, M. (2023) Model-robust inference for clinical trials that improve precision by stratified randomization and covariate adjustment. Journal of the American Statistical Association, 118, 1152--1163

    2023

  36. [44]

    (1978) An application of an urn model to the design of sequential controlled clinical trials

    Wei, L.J. (1978) An application of an urn model to the design of sequential controlled clinical trials. Journal of the American Statistical Association, 72, 382--386

    1978

  37. [45]

    and Zhao, Q

    Ye, T., Shao, J., Yi, Y. and Zhao, Q. (2023) Toward better practice of covariate adjustment in analyzing randomized clinical trials. Journal of the American Statistical Association, 118, 2370--2382

    2023

  38. [46]

    and Yang, Y

    Liu, H. and Yang, Y. (2020) Regression-adjusted average treatment effect estimates in stratified randomized experiments. Biometrika, 107, 935--948

    2020

  39. [47]

    (1996) A Course in Large Sample Theory

    Ferguson, T.S. (1996) A Course in Large Sample Theory. Boca Raton, Florida, Chapman and Hall

    1996

  40. [48]

    (2003) Mathematical Statistics, 2nd edition

    Shao, J. (2003) Mathematical Statistics, 2nd edition. New York: Springer

    2003

  41. [49]

    van der Vaart, A. W. (1998) Asymptotic Statistics. Cambridge, UK: Cambridge University Press

    1998

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.