Pith. sign in

REVIEW 5 minor 74 references

When treatments are random but covariates are fixed, usual robust standard errors stay valid but are conservative, and common regressions still recover average treatment effects under misspecification.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-13 02:18 UTC pith:2H3RQWPC

load-bearing objection Clean mixed-design Z-estimation theory that fills a real gap between random and fixed regressors, with useful causal interpretations and a novel cluster variance correction.

arxiv 2607.09536 v1 pith:2H3RQWPC submitted 2026-07-10 math.ST econ.EMstat.MEstat.TH

Misspecified regressions with mixed regressors: robust inference and causal interpretation

Mengsi Gao , Peng Ding This is my paper
classification math.ST econ.EMstat.MEstat.TH MSC 62J0562F1262K99
keywords mixed regressorsmisspecificationHuber-White standard erroraverage treatment effectregression adjustmentcluster randomizationZ-estimationcausal inference
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Standard regression theory treats all regressors as either fully random or fully fixed. That split leaves out the everyday experimental design in which treatment is randomized while pretreatment covariates are fixed. This paper fills the gap by developing asymptotic theory for estimating equations and least-squares regressions that mix random and fixed regressors, even when the working model is wrong. The central result is that the usual sandwich (Huber–White or Liang–Zeger) variance estimator remains asymptotically valid but is generally conservative once some regressors are conditioned on. Applied to completely randomized and cluster-randomized experiments, the same theory shows that difference-in-means, Fisher’s additive adjustment, and Lin’s fully interacted adjustment all target well-defined average treatment effects (or their covariate-conditional versions), and that a simple correction restores consistency of the variance estimator under fully random covariates. The practical payoff is clear guidance on which regression coefficients and standard errors can be trusted for causal inference when models are misspecified and designs are mixed.

Core claim

Under a mixed design that conditions only on a subset of the regressors, the Z-estimator (and its OLS special case) is asymptotically normal for the mixed estimand, and the ordinary sandwich variance estimator converges to the true asymptotic variance plus a nonnegative bias term given by the outer product of the conditional means of the estimating equations. Consequently the sandwich is conservative, and it becomes exact precisely when those conditional means vanish.

What carries the argument

The mixed-design sandwich decomposition: nV̂hw = Vm + Bm + oP(1), where Bm is the average outer product of the estimating-equation residuals conditional on the fixed regressors. This identity simultaneously yields asymptotic normality of the estimator and the precise sense in which robust standard errors overstate uncertainty.

Load-bearing premise

The fourth-moment and positive-definiteness conditions that justify the law of large numbers and central limit theorem for the estimating equations; if those moments fail, the normality and conservativeness claims do not hold as stated.

What would settle it

Simulate a completely randomized experiment with fixed covariates and a deliberately misspecified linear model, compute the usual Huber–White standard error and the Monte-Carlo standard deviation of the treatment coefficient; if the sandwich is systematically smaller than the Monte-Carlo SD, the conservativeness claim is false.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 5 minor

Summary. The paper develops asymptotic theory for Z-estimation (and OLS as a special case) under three designs—random, fixed, and mixed regressors—allowing for misspecification. The central contribution is the mixed-design case: conditioning on a subset of regressors X2 while treating the rest as random. Under standard moment conditions, the Z-estimator is asymptotically normal for the mixed estimand βm, and the usual Huber–White (or Liang–Zeger) sandwich is conservative, nV̂hw = Vm + Bm + oP(1) with Bm ≽ 0 given by the outer product of the conditional means of the estimating equations (Theorems 2.3, 3.3, 5.3). The theory is specialized to completely randomized and cluster-randomized experiments, clarifying the causal meaning of Fisher and Lin regression coefficients under misspecification and providing a novel correction to the EHW/LZ variance for fully interacted (Lin) regressions under random design (Theorems 4.4, 5.7). Simulations match the predicted consistency/conservativeness/anti-conservativeness patterns.

Significance. The mixed-design framework fills a genuine gap between pure random- and fixed-regressor theory and matches the common experimental practice of randomized treatment with fixed pretreatment covariates. The conservativeness result for sandwich SEs under mixed design, the causal interpretation of misspecified Fisher/Lin coefficients, and the novel variance correction for Lin-type regressions under random design are useful and carefully derived. The paper supplies detailed proofs in the supplement and Monte Carlo evidence (n=1000, M=160, B=5000) that tracks the theory. If the results hold as stated, they give applied researchers a coherent justification for using (and, when needed, correcting) robust SEs in regression-adjusted RCTs and cluster RCTs under misspecification.

minor comments (5)
  1. Table 1 is a helpful overview; consider adding a short note in the caption that “anti-conservative” for Lin under random design is resolved by the adjusted SE in Theorems 4.4 and 5.7, so readers do not misread the table as recommending against Lin.
  2. In §4.3 and §5.2.3 the two constructions of the corrected variance (augmented estimating equations vs. adding the correction term to the usual EHW/LZ) are said to be asymptotically equivalent but not numerically identical. A one-sentence remark on which is preferred in practice would help applied readers.
  3. Assumption 5(g) (Ω = o(M−2/3)) is used for the cluster Lin results; a brief pointer to where this rate is sharp (or whether a weaker rate would suffice for the main conservativeness claim) would improve transparency.
  4. The IV/LATE results are relegated to the supplement; a short forward reference in the main text (e.g., at the end of §4) would make the scope of the paper clearer without expanding the main body.
  5. Notation for the mixed estimand βm and the bias Bm is consistent, but the switch between E◦ and the design-specific operators is dense in §2; a small display summarizing the three (⋄, ◦) pairs would aid readability.

Circularity Check

0 steps flagged

No significant circularity: mixed-design sandwich conservativeness is derived by expansion of estimating equations, not by fitting or self-definition.

full rationale

The paper's central claims (Theorems 2.3, 3.3, 5.3 and the RCT/cluster specializations) start from population or conditional moment conditions that define the estimands βr, βf, βm, then obtain asymptotic normality by standard Taylor linearization of the sample estimating equations and obtain the sandwich bias terms Bm (and Bf) by decomposing the middle matrix into conditional variance plus the outer product of conditional means of the scores. Those bias terms are not free parameters fitted to the target; they are explicit functions of the same moments that define the estimands. The novel random-design correction for Lin-type fully interacted regressions (adding the (γ̂₁−γ̂₀)ᵀΣx(γ̂₁−γ̂₀) term) is likewise obtained by augmenting the Z-estimation system with the moment for the sample mean of covariates, not by renaming a fit. Self-citations (Abadie et al. 2014, Lin 2013, Negi & Wooldridge 2021, Su & Ding 2021, etc.) are used as benchmarks or for known special cases; none of them is load-bearing for the mixed-design conservativeness result, which is proved from first principles under Assumptions 1–5. No step reduces the claimed prediction to its own input by construction. Score 1 only for ordinary, non-load-bearing self-citation of related design-based work.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

The paper is pure asymptotic theory under standard regularity conditions; it introduces no free parameters fitted to data and no new physical or statistical entities. All load-bearing ingredients are either classical probability tools or domain assumptions standard in causal inference and cluster asymptotics.

axioms (4)
  • standard math Standard LLN, CLT (including Lindeberg–Feller conditional versions) and continuous-mapping/Slutsky arguments for Z-estimators under the stated moment conditions.
    Used throughout Sections 2–5 and the supplement to obtain asymptotic normality and consistency/conservativeness of sandwich estimators.
  • domain assumption i.i.d. sampling of units (or independent clusters) with finite (2+δ) or fourth moments on outcomes and regressors, and nonsingularity of the Jacobian Γ.
    Assumptions 1–5; required for the uniform LLN and Lyapunov conditions that deliver the oP(1) remainders.
  • domain assumption Randomization of treatment (Zi ⊥ potential outcomes and covariates) with fixed propensity e ∈ (0,1); for LATE, monotonicity (no defiers).
    Assumptions 3 and 6; needed for the regression coefficients to identify ATE/LATE rather than mere projection coefficients.
  • domain assumption For clustered data, maximum relative cluster size Ω = o(M−2/3).
    Assumption 5(g); used to control remainder terms in the cluster-level expansions (Lemmas S5.1–S5.5).

pith-pipeline@v1.1.0-grok45 · 86738 in / 2514 out tokens · 38368 ms · 2026-07-13T02:18:16.413609+00:00 · methodology

0 comments
read the original abstract

For analytic convenience, existing statistical frameworks either assume random or fixed regressors. However, it is a little awkward that they do not cover the practical case of estimating the average treatment effect in experiments with randomized treatments and non-randomized, fixed pretreatment covariates. We unify the literature by providing the theory for regressions with mixed regressors that contain both random and fixed components. Importantly, our theory allows for misspecification of the regression functions. We first establish general results for estimating equations with both random and fixed components and then use it to analyze misspecified linear regression, with applications to completely randomized experiments. We focus on the causal interpretation of the regression coefficients and standard errors even when the models are wrong. We start with the theory for independent data and then extend the discussion to clustered data.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

74 extracted references · 23 canonical work pages · 5 internal anchors

  1. [1]

    and Wooldridge, Jeffrey M

    Abadie, Alberto and Athey, Susan and Imbens, Guido W. and Wooldridge, Jeffrey M. , year = 2020, journal =. Sampling-. doi:10.3982/ECTA12675 , urldate =

    doi:10.3982/ecta12675 2020

  2. [2]

    Abadie, Alberto and Athey, Susan and Imbens, Guido W and Wooldridge, Jeffrey M , year = 2023, month = feb, journal =. When. doi:10.1093/qje/qjac038 , urldate =

    doi:10.1093/qje/qjac038 2023

  3. [3]

    , year = 2008, journal =

    Abadie, Alberto and Imbens, Guido W. , year = 2008, journal =. On the. doi:10.3982/ECTA6474 , urldate =

    doi:10.3982/ecta6474 2008

  4. [4]

    and Zheng, Fanyin , year = 2014, month = oct, journal =

    Abadie, Alberto and Imbens, Guido W. and Zheng, Fanyin , year = 2014, month = oct, journal =. Inference for. doi:10.1080/01621459.2014.928218 , urldate =

    doi:10.1080/01621459.2014.928218 2014

  5. [5]

    The Purpose of an Estimator Is What It Does:

    Andrews, Isaiah and Chen, Jiafeng and Tecchio, Otavio , year = 2025, month = aug, number =. The Purpose of an Estimator Is What It Does:. doi:10.48550/arXiv.2508.13076 , urldate =. arXiv , keywords =:2508.13076 , primaryclass =

    doi:10.48550/arxiv.2508.13076 2025

  6. [6]

    Quantile

    Angrist, Joshua and Chernozhukov, Victor and. Quantile. Econometrica , volume =. doi:10.1111/j.1468-0262.2006.00671.x , urldate =

    doi:10.1111/j.1468-0262.2006.00671.x 2006

  7. [7]

    and Imbens, Guido W

    Angrist, Joshua D. and Imbens, Guido W. and Rubin, Donald B. , year = 1996, journal =. Identification of. doi:10.2307/2291629 , urldate =. 2291629 , eprinttype =

    doi:10.2307/2291629 1996

  8. [8]

    Angrist, Joshua and Pischke, J. Mostly. Mostly

  9. [9]

    , year = 1987, journal =

    Arellano, M. , year = 1987, journal =. doi:10.1111/j.1468-0084.1987.mp49004006.x , urldate =

    work page doi:10.1111/j.1468-0084.1987.mp49004006.x 1987

  10. [10]

    and Koles

    Armstrong, Timothy B. and Koles. Finite-. Econometrica , volume =. doi:10.3982/ECTA16907 , urldate =

    work page doi:10.3982/ecta16907

  11. [11]

    and Koles

    Barrios, Thomas and Diamond, Rebecca and Imbens, Guido W. and Koles. Clustering,. Journal of the American Statistical Association , volume =. doi:10.1080/01621459.2012.682524 , urldate =

    doi:10.1080/01621459.2012.682524 2012

  12. [12]

    Box, George E. P. and Draper, Norman R. , year = 1987, month = jan, publisher =. Empirical

    1987

  13. [13]

    and Canay, Ivan A

    Bugni, Federico A. and Canay, Ivan A. and Shaikh, Azeem M. , year = 2018, month = oct, journal =. Inference. doi:10.1080/01621459.2017.1375934 , urldate =

    doi:10.1080/01621459.2017.1375934 2018

  14. [14]

    Quantitative Economics , volume =

    Inference under Covariate-Adaptive Randomization with Multiple Treatments , author =. Quantitative Economics , volume =. doi:10.3982/QE1150 , urldate =

    doi:10.3982/qe1150

  15. [15]

    Inference for Cluster Randomized Experiments with Non-ignorable Cluster Sizes

    Bugni, Federico and Canay, Ivan and Shaikh, Azeem and. Inference for. doi:10.48550/arXiv.2204.08356 , urldate =. arXiv , keywords =:2204.08356 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2204.08356

  16. [16]

    and Shaikh, Azeem M

    Bugni, Federico and Canay, Ivan A. and Shaikh, Azeem M. and. Inference for. Journal of Political Economy Microeconomics , volume =. doi:10.1086/732836 , urldate =

    doi:10.1086/732836

  17. [17]

    Journal of Econometrics , volume =

    Inference under Covariate-Adaptive Randomization with Imperfect Compliance , author =. Journal of Econometrics , volume =. doi:10.1016/j.jeconom.2023.105497 , urldate =

    doi:10.1016/j.jeconom.2023.105497 2023

  18. [18]

    Statistics for Big Data:

    B. Statistics for Big Data:. Statistics & Probability Letters , volume =

  19. [19]

    Models as

    Buja, Andreas and Brown, Lawrence and Berk, Richard and George, Edward and Pitkin, Emil and Traskin, Mikhail and Zhang, Kai and Zhao, Linda , year = 2019, journal =. Models as. 26890741 , eprinttype =

    2019

  20. [20]

    Models as

    Buja, Andreas and Brown, Lawrence and Kuchibhotla, Arun Kumar and Berk, Richard and George, Edward and Zhao, Linda , year = 2019, journal =. Models as. 26890742 , eprinttype =

    2019

  21. [21]

    and Gertler, Paul J

    Celhay, Pablo A. and Gertler, Paul J. and Giovagnoli, Paula and Vermeersch, Christel , year = 2019, month = jul, journal =. Long-. doi:10.1257/app.20170128 , urldate =

    work page doi:10.1257/app.20170128 2019

  22. [22]

    Standard errors when a regressor is randomly assigned

    Standard Errors When a Regressor Is Randomly Assigned , author =. doi:10.48550/arXiv.2303.10306 , urldate =. arXiv , keywords =:2303.10306 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2303.10306

  23. [23]

    Economic Modelling , volume =

    Maximum-Likelihood Estimation of Misspecified Models , author =. Economic Modelling , volume =

  24. [24]

    , year = 2003, month = oct, journal =

    Christopeit, N. , year = 2003, month = oct, journal =. Wooldridge,. doi:10.1007/s00712-003-0589-6 , urldate =

    doi:10.1007/s00712-003-0589-6 2003

  25. [25]

    Conley, T. G. , year = 1999, month = sep, journal =. doi:10.1016/S0304-4076(98)00084-0 , urldate =

    doi:10.1016/s0304-4076(98)00084-0 1999

  26. [26]

    Cox, D. R. , year = 1961, month = jan, volume =. Tests of. Proceedings of the

    1961

  27. [27]

    Two Seemingly Paradoxical Results in Linear Models: The Variance Inflation Factor and the Analysis of Covariance , shorttitle =

    Ding, Peng , year = 2021, journal =. Two Seemingly Paradoxical Results in Linear Models: The Variance Inflation Factor and the Analysis of Covariance , shorttitle =

    2021

  28. [28]

    Ding, Peng , year = 2021, month = jan, journal =. The. doi:10.1016/j.spl.2020.108945 , urldate =

    doi:10.1016/j.spl.2020.108945 2021

  29. [29]

    Ding, Peng , year = 2023, month = dec, number =. Linear. doi:10.48550/arXiv.2401.00649 , urldate =. arXiv , keywords =:2401.00649 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2401.00649 2023

  30. [30]

    Ding, Peng , year = 2024, month = jul, publisher =. A. doi:10.1201/9781003484080 , abstract =

    work page doi:10.1201/9781003484080 2024

  31. [31]

    Decomposing

    Ding, Peng and Feller, Avi and Miratrix, Luke , year = 2019, month = jan, journal =. Decomposing. doi:10.1080/01621459.2017.1407322 , urldate =

    doi:10.1080/01621459.2017.1407322 2019

  32. [32]

    Limit Theorems for Regressions with Unequal and Dependent Errors , booktitle =

    Eicker, Friedhelm , year = 1967, month = jan, volume =. Limit Theorems for Regressions with Unequal and Dependent Errors , booktitle =

    1967

  33. [33]

    , year = 2014, month = dec, journal =

    Feng, Changyong and Wang, Hongyue and Chen, Tian and Tu, Xin M. , year = 2014, month = dec, journal =. On Exact Forms of. doi:10.1093/biomet/asu061 , urldate =

    work page doi:10.1093/biomet/asu061 2014

  34. [34]

    The Design of Experiments , author =

  35. [35]

    Freedman, David A , year = 2006, month = nov, journal =. On. doi:10.1198/000313006X152207 , urldate =

    work page doi:10.1198/000313006x152207 2006

  36. [36]

    Advances in Applied Mathematics , volume =

    On Regression Adjustments to Experimental Data , author =. Advances in Applied Mathematics , volume =. doi:10.1016/j.aam.2006.12.003 , urldate =

    doi:10.1016/j.aam.2006.12.003 2006

  37. [37]

    Discussion of

    van de Geer, Sara , year = 2019, month = nov, journal =. Discussion of. doi:10.1214/19-STS722 , urldate =

    work page doi:10.1214/19-sts722 2019

  38. [38]

    and Lee, Seojeong , year = 2019, month = jun, journal =

    Hansen, Bruce E. and Lee, Seojeong , year = 2019, month = jun, journal =. Asymptotic. doi:10.1016/j.jeconom.2019.02.001 , urldate =. arXiv , keywords =:1902.01497 , primaryclass =

    work page doi:10.1016/j.jeconom.2019.02.001 2019

  39. [39]

    arXiv.org , urldate =

    Peer Effects in the Linear-in-Means Model May Be Inestimable Even When Identified , author =. arXiv.org , urldate =

  40. [40]

    , year = 1967, month = jan, volume =

    Huber, Peter J. , year = 1967, month = jan, volume =. The Behavior of Maximum Likelihood Estimates under Nonstandard Conditions , booktitle =

    1967

  41. [41]

    , year = 1997, month = jul, journal =

    Imbens, Guido W. , year = 1997, month = jul, journal =. One-. doi:10.2307/2971718 , urldate =

    work page doi:10.2307/2971718 1997

  42. [42]

    and Angrist, Joshua D

    Imbens, Guido W. and Angrist, Joshua D. , year = 1994, journal =. Identification and. doi:10.2307/2951620 , urldate =. 2951620 , eprinttype =

    doi:10.2307/2951620 1994

  43. [43]

    Tailored Inference for Finite Populations: Conditional Validity and Transfer across Distributions , shorttitle =

    Jin, Ying and Rothenh. Tailored Inference for Finite Populations: Conditional Validity and Transfer across Distributions , shorttitle =. Biometrika , pages =. doi:10.1093/biomet/asad022 , urldate =

    work page doi:10.1093/biomet/asad022

  44. [44]

    Model-free Study of Ordinary Least Squares Linear Regression

    Kuchibhotla, Arun K. and Brown, Lawrence D. and Buja, Andreas , year = 2018, month = sep, number =. Model-Free. doi:10.48550/arXiv.1809.10538 , urldate =. arXiv , keywords =:1809.10538 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1809.10538 2018

  45. [45]

    Biometrika , volume =

    Longitudinal Data Analysis Using Generalized Linear Models , author =. Biometrika , volume =. doi:10.1093/biomet/73.1.13 , urldate =

    doi:10.1093/biomet/73.1.13

  46. [46]

    Bayesian Causal Inference: A Critical Review , shorttitle =

    Li, Fan and Ding, Peng and Mealli, Fabrizia , year = 2023, month = mar, journal =. Bayesian Causal Inference: A Critical Review , shorttitle =. doi:10.1098/rsta.2022.0153 , urldate =

    doi:10.1098/rsta.2022.0153 2023

  47. [47]

    Agnostic Notes on Regression Adjustments to Experimental Data:

    Lin, Winston , year = 2013, month = mar, journal =. Agnostic Notes on Regression Adjustments to Experimental Data:. doi:10.1214/12-AOAS583 , urldate =

    doi:10.1214/12-aoas583 2013

  48. [48]

    Robust and Efficient Estimation of Potential Outcome Means under Random Assignment

    Negi, Akanksha and Wooldridge, Jeffrey M. , year = 2020, month = oct, number =. Robust and. doi:10.48550/arXiv.2010.01800 , urldate =. arXiv , keywords =:2010.01800 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2010.01800 2020

  49. [49]

    Econometric Reviews , volume =

    Revisiting Regression Adjustment in Experiments with Heterogeneous Treatment Effects , author =. Econometric Reviews , volume =. doi:10.1080/07474938.2020.1824732 , urldate =

    work page doi:10.1080/07474938.2020.1824732 2020

  50. [50]

    Economics Letters , volume =

    A Method of Moments Interpretation of Sequential Estimators , author =. Economics Letters , volume =

  51. [51]

    and McFadden, Daniel , year = 1994, month = jan, volume =

    Newey, Whitney K. and McFadden, Daniel , year = 1994, month = jan, volume =. Chapter 36. Handbook of. doi:10.1016/S1573-4412(05)80005-4 , urldate =

    doi:10.1016/s1573-4412(05)80005-4 1994

  52. [52]

    and West, Kenneth D

    Newey, Whitney K. and West, Kenneth D. , year = 1987, journal =. A. doi:10.2307/1913610 , urldate =. 1913610 , eprinttype =

    doi:10.2307/1913610 1987

  53. [53]

    Neyman, Jerzy , year = 1923, journal =. On the. doi:10.1214/ss/1177012031 , urldate =

    doi:10.1214/ss/1177012031 1923

  54. [54]

    Ren, Jiyang , year = 2024, month = apr, journal =. Model-. doi:10.1080/07350015.2023.2224851 , urldate =

    doi:10.1080/07350015.2023.2224851 2024

  55. [55]

    , year = 2020, month = jan, journal =

    Rosset, Saharon and Tibshirani, Ryan J. , year = 2020, month = jan, journal =. From. doi:10.1080/01621459.2018.1424632 , urldate =

    doi:10.1080/01621459.2018.1424632 2020

  56. [56]

    Statistics & Probability Letters , volume =

    On Equivalencies between Design-Based and Regression-Based Variance Estimators for Randomized Experiments , author =. Statistics & Probability Letters , volume =. doi:10.1016/j.spl.2011.10.024 , urldate =

    work page doi:10.1016/j.spl.2011.10.024 2011

  57. [57]

    , year = 2013, journal =

    Schochet, Peter Z. , year = 2013, journal =. Estimators for. 41999423 , eprinttype =

    2013

  58. [58]

    Shen, Dennis and Ding, Peng and Sekhon, Jasjeet and Yu, Bin , year = 2023, journal =. Same. doi:10.3982/ECTA21248 , urldate =

    work page doi:10.3982/ecta21248 2023

  59. [59]

    doi:10.48550/arXiv.2411.00947 , urldate =

    Asymptotic Theory of the Quadratic Assignment Procedure for Dyadic Data Analysis , author =. doi:10.48550/arXiv.2411.00947 , urldate =. arXiv , keywords =:2411.00947 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2411.00947

  60. [60]

    , year = 2010, month = jun, journal =

    Sims, Christopher A. , year = 2010, month = jun, journal =. But. doi:10.1257/jep.24.2.59 , urldate =

    work page doi:10.1257/jep.24.2.59 2010

  61. [61]

    Su, Fangzhou and Ding, Peng , year = 2021, month = nov, journal =. Model-. doi:10.1111/rssb.12468 , urldate =

    work page doi:10.1111/rssb.12468 2021

  62. [62]

    van der Vaart, A. W. , year = 1998, publisher =. Asymptotic

    1998

  63. [63]

    van der Vaart, A. W. , year = 2000, month = jun, publisher =. Asymptotic

    2000

  64. [64]

    and Li, Fan , year = 2024, month = oct, journal =

    Wang, Bingkai and Park, Chan and Small, Dylan S. and Li, Fan , year = 2024, month = oct, journal =. Model-. doi:10.1080/01621459.2023.2289693 , urldate =

    doi:10.1080/01621459.2023.2289693 2024

  65. [65]

    White, Halbert , year = 1980, journal =. A. doi:10.2307/1912934 , urldate =. 1912934 , eprinttype =

    doi:10.2307/1912934 1980

  66. [66]

    White, Halbert , year = 1980, journal =. Using

    1980

  67. [67]

    White, Halbert , year = 1982, journal =. Maximum. doi:10.2307/1912526 , urldate =. 1912526 , eprinttype =

    doi:10.2307/1912526 1982

  68. [68]

    Corrigendum [

    White, Halbert , year = 1983, journal =. Corrigendum [

    1983

  69. [69]

    , year = 2020, publisher =

    Wooldridge, Jeffrey M. , year = 2020, publisher =. Introductory

    2020

  70. [70]

    Potential Outcomes and Finite-Population Inference for

    Xu, Ruonan , year = 2021, month = jan, journal =. Potential Outcomes and Finite-Population Inference for. doi:10.1093/ectj/utaa022 , urldate =

    work page doi:10.1093/ectj/utaa022 2021

  71. [71]

    Covariate-Adjusted

    Zhao, Anqi and Ding, Peng , year = 2021, month = dec, journal =. Covariate-Adjusted. doi:10.1016/j.jeconom.2021.04.007 , urldate =

    doi:10.1016/j.jeconom.2021.04.007 2021

  72. [72]

    The Annals of Statistics , volume =

    Reconciling Design-Based and Model-Based Causal Inferences for Split-Plot Experiments , author =. The Annals of Statistics , volume =. doi:10.1214/21-AOS2144 , urldate =

    work page doi:10.1214/21-aos2144

  73. [73]

    Covariate adjustment in randomized experiments with missing outcomes and covariates

    Covariate Adjustment in Randomized Experiments with Missing Outcomes and Covariates , author =. doi:10.48550/arXiv.2311.10877 , urldate =. arXiv , keywords =:2311.10877 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2311.10877

  74. [75]

    doi:10.48550/arXiv.2502.00251 , urldate =

    Two-Stage Least Squares with Treatment-Covariate Interactions for Treatment Effect Heterogeneity , author =. doi:10.48550/arXiv.2502.00251 , urldate =. arXiv , keywords =:2502.00251 , primaryclass =

    work page doi:10.48550/arxiv.2502.00251