Pith. sign in

REVIEW 3 major objections 5 minor 141 references

Treating an estimate as a function on the parameter space, not a point, yields a uniform optimum the classical theory could not get.

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 ↗

Generalized estimators (sample-to-function maps) admit a uniform information bound attained by the score, explaining classical estimation strains without overturning its theorems.

T0 review reviewed 2026-07-31 challenge →

load-bearing objection Clean Bahadur reread that reframes classical strains as features; math is elementary and mostly right, with one standard regularity gap that is fixable. the 3 major comments →

arxiv 2607.24890 v1 pith:2HU2SSY2 submitted 2026-07-27 math.ST cs.ITmath.ITstat.TH

A New Look at the Classical Estimation Problem

Paul W. Vos This is my paper
classification math.ST cs.ITmath.ITstat.TH MSC 62F1062B1062A01
keywords Bahadur's lecturesgeneralized estimationFisher informationscore functionunbiased estimationsufficiencyCramér–Raopointwise risk
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.

The reading

Classical point estimation has no uniformly best estimator under any risk that depends on one distribution at a time; that structural gap is why the field invented unbiasedness, admissibility, minimaxity, and Bayes averaging. This paper keeps Bahadur's Hilbert-space development and changes only the object: at each sample the estimate becomes a smooth function on the whole parameter space—the continuum of null hypotheses Fisher described—rather than a single number. The new object is assessed by how fast it responds as the true distribution moves, a family-aware information measure Λ. Under that measure the score attains the Fisher information bound at every parameter value by a three-line argument, with no side constraints. Classical strains—missing unbiased estimates, absurd two-point solutions, maximum-likelihood heuristics, boundary samples—become exact statements or non-problems, while every classical theorem remains true.

Core claim

No pointwise risk admits a uniformly optimal point estimator. Once an estimator is redefined as a centered, smooth function τ on sample space times parameter space, its information Λ(τ) equals the squared correlation of τ with the score times Fisher information I, hence is at most I, with equality precisely when τ points along the score. The score therefore attains the bound uniformly, recovering Cramér–Rao attainment and sufficiency as equality cases and converting the maximum-likelihood heuristics into exact statements about the score.

What carries the argument

The information Λ(τ) = [E(τ′)]² / Var(τ) of a generalized estimator, together with the differentiation identity that turns E(τ′) into −E(score · τ). Cauchy–Schwarz then yields the uniform bound in three lines.

Load-bearing premise

The estimate must be continuously differentiable in the parameter and expectations must be differentiable under the integral; without that smoothness the information measure is undefined and the short optimality proof does not run.

What would settle it

Exhibit a regular one-parameter family and a generalized estimator whose Λ exceeds Fisher information at even one parameter value, or whose correlation with the score is one yet is not a multiple of the score.

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

If this is right

  • Boundary samples (zero successes, odds-ratio plug-in infinite) still yield usable standardized-score curves and intervals.
  • Locally best unbiased estimates, kept with their parameter dependence, assemble into one fully efficient generalized estimator.
  • Higher-order Bhattacharya spans, admissibility, minimaxity and unbiasedness become optional apparatus once the object is family-aware.
  • Sufficiency and uniform Cramér–Rao attainment are equality cases of one information bound under two maps from ordinary estimators.
  • Maximum-likelihood heuristics become exact first-order statements that the score is already optimal.

Where Pith is reading between the lines

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

  • The same continuum-of-tests reading suggests multiparameter and nuisance-parameter extensions via successive orthogonalization of unwanted directions.
  • A criterion that can separate a potential function (e.g., log-likelihood) from its derivative would answer the open question the paper leaves about one-sided procedures.
  • Nonsmooth families excluded here could be recovered by replacing ordinary derivatives in Λ with difference quotients in the style of Chapman–Robbins.
Share X Bluesky LinkedIn Reddit HN

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

3 major / 5 minor

Summary. The paper re-presents a subset of Bahadur's Lectures on the Theory of Estimation with one modification: an estimator is a function τ on S×Θ, mean-zero under P_θ at each θ, rather than a point-valued statistic. Within this framework the author (i) proves an elementary impossibility lemma (Lemma 1) showing no pointwise risk admits a uniformly optimal estimator, and reads admissibility, minimaxity, Bayes, and unbiasedness as responses to it; (ii) defines an information criterion Λ(τ)=(Eτ′)²/Var(τ) and proves (Theorem 1) that Λ(τ)=Corr²(τ,γ⁽¹⁾)I≤I with equality iff τ is a scalar multiple of the score; (iii) recovers Cramér–Rao attainment and sufficiency as equality cases under two maps from point estimators to generalized estimators; and (iv) re-reads Bahadur's ML heuristics as exact statements about the score. The mathematics is deliberately elementary — Hilbert-space geometry and one differentiation identity (7) — and the presentation follows Bahadur's notation closely with a running Bernoulli example.

Significance. If the regularity gap noted below is closed, the paper delivers what it promises: a self-contained, parameter-free development in which the score's optimality is a three-line consequence of Cauchy–Schwarz (Theorem 1), the absence of uniformly optimal point estimators is proved in four lines without any loss-function choice (Lemma 1), and sufficiency and CR attainment fall out as equality cases rather than being assumed. These are genuinely reusable pedagogical and conceptual contributions: the unification of estimation and testing under one object, and the exact finite-sample reading of Efron's curvature via Eff(τ_θ̂)=Corr²(τ_θ̂,γ⁽¹⁾) in §5.3, are stated crisply and without asymptotics. The explicit scope limits (scalar parameter, smooth case, no asymptotics) and the honest accounting of where Λ is silent (§8, potential functions) strengthen the paper. The individual ingredients are classical — the bound is Cauchy–Schwarz and the connection to estimating-function optimality (Godambe–Thompson) is acknowledged in §6.3 — so the novelty is organizational and interpretive rather than technical; at that level the paper is well executed and worth publishing once the hypotheses match the量化

major comments (3)
  1. [§4.2, Theorem 1; §5.1] Theorem 1 (via Eq. (11) and Definition 3): the quantifier "every generalized estimator" is not supported by the hypotheses. Identity (7) is legitimated in §5.1 by Conditions 1–3 for integrands whose θ-dependence is only through the measure (Condition 2 dominates the score m_θ, not the integrand). But (11) applies (7) to f(s,θ)=τ(s,θ), which depends on θ directly, and nothing in Definition 2 or Conditions 1–3 dominates ∂_θτ or ensures uniform integrability of the difference quotients of E_θτ(·,θ). Definition 2 gives τ(·,θ)∈L²(P_θ) pointwise and τ(s,·) C¹ a.e.; this does not imply ∂_θτ(·,θ)∈L¹(P_θ), so E(τ′) in Definition 3 may fail to exist for objects satisfying (i)–(iii). The fix is standard — add to Definition 2(ii) a local domination such as E_θ[sup_{|δ−θ|≤ε}|∂_θτ(s,δ)|]<∞ or a uniform-integrability condition — and verify it for the running examples (τ_E, τ_t, the score). Relatedly, t
  2. [§3.3, Eq. (6)] The assembled LMVUE τ(s,θ)=t̃(s)−g(θ) is asserted to be a generalized estimator "granted the smoothness in θ that Bahadur's next chapter assumes anyway." But Conditions 1–3 concern smoothness of ℓ_θ, not of the projection map θ↦π_{W_θ}t. That the LMVUE at θ varies C¹ in θ is an additional hypothesis (essentially smoothness of the family of subspaces W_θ), used again in §4.2 where the first-order version g′(θ)γ⁽¹⁾/I is declared fully efficient. This should be stated as an explicit assumption or proved under named conditions, since it feeds the paper's central interpretive claim that "the problem in practice" is the solution.
  3. [§5.3, after Eq. (13)] The equivalence "I_t=I for every θ iff t is sufficient — equivalently τ_t=γ⁽¹⁾" is load-bearing for the claim that sufficiency is recovered as an equality case, but it is stated without proof, citation, or regularity conditions. The forward direction (sufficiency preserves information) is standard; the converse (I_t=I everywhere implies sufficiency) is a real theorem requiring hypotheses — a.s. equality γ⁽¹⁾=γ⁽¹⁾_t∘t for all θ yields likelihood ratios depending on s only through t under positivity and mutual absolute continuity — and should either be proved in the paper's Hilbert-space idiom or given a precise citation with conditions. As written the reader cannot tell which of Conditions 1–3, or what additional assumptions, are in force.
minor comments (5)
  1. [§4.2, Definition 4 / Corollary 1] Corr(τ,γ⁽¹⁾) and Eff(τ) are functions of θ (computed under P_θ) but are written without an argument; stating Eff(τ)(θ) once in Definition 4/Corollary 1 would prevent misreading, especially where "uniformly" and "at every θ" are quantified.
  2. [§4.2, final paragraph] The claim that among generalized estimators that are functions of t the marginal score τ_t maximizes Corr with γ⁽¹⁾ is attributed to the author's preprint [Vos, 2026]. Since this is a clean projection statement (the marginal score is the conditional expectation of γ⁽¹⁾ given t, and conditioning is the L² projection), a two-line self-contained proof would make the paper independent of an unpublished source.
  3. [§1; §4.3] Footnote 1 marker appears inside "Bhattacharya1 bounds" in the introduction, rendering as "Bhattacharya1"; check typesetting. Also plan (c) of the running example is introduced "for fidelity to the source but not discussed further" — consider deleting it, since its only effect is to make (a),(b),(c) labeling look like an omission.
  4. [§1] The Fisher (1955) quotation carrying the paper's motivating interpretation ("a continuum of hypotheses each eligible as null hypothesis") is from a general-discussion paper; a sentence noting the context in which Fisher wrote it, and perhaps Fisher (1922) §4 on information, would help readers weigh how much interpretive load the quote can bear.
  5. [§2.2] In §2.2 the "invariance" paragraph says the standardization τ̄ is "the same function on the family however the family is labeled"; this is true because both τ and its variance transform covariantly, but the one-line justification given (relabeling carries τ̄ to the corresponding function) is circular as an explanation. A half-sentence showing Var transforms by the squared derivative would close it.

Circularity Check

1 steps flagged

No significant circularity: Theorem 1 and Lemma 1 are self-contained Hilbert-space arguments; self-citations supply framework context only.

specific steps
  1. self citation load bearing [Section 1, paragraph on framework sources; also Sections 5.3–5.4, 8]
    "The framework presented here is developed in Vos and Wu [2025] in the language of information geometry, and applied to the James–Stein paradox in Vos [2026]. Neither source is presupposed."

    Overlapping-author citations supply the broader multiparameter geometry and applications. They are not load-bearing for Theorem 1 or Lemma 1, which are proved in-paper from Bahadur plus Definition 2; flagged only as minor self-citation context, not as a reduction of the central claim.

full rationale

The paper’s load-bearing results do not reduce to their inputs by construction. Lemma 1 is an elementary contradiction from common support and non-constant g; its proof uses only the definition of a pointwise risk. Theorem 1 defines Λ(τ) = [E(τ′)]²/Var(τ), applies the classical differentiation identity (7) to obtain the score equation E(τ′) = −E(γ⁽¹⁾τ), and finishes by Cauchy–Schwarz—three lines that do not smuggle the bound into the definition. Cramér–Rao, sufficiency, and ML heuristics are recovered as equality cases or re-readings under two explicit maps from point estimators to generalized estimators; the classical statements remain theorems and are not renamed empirical patterns. Self-citations to Vos & Wu (2025) and Vos (2026) appear for the multiparameter/information-geometry setting and the James–Stein application, but the paper states that neither source is presupposed and re-derives the scalar theory from Bahadur’s lectures plus Definition 2. No parameter is fitted to data and then called a prediction; no uniqueness theorem is imported to forbid alternatives. The skeptic’s regularity gap (domination for ∂_θτ) is a correctness/scope issue, not circularity. Score 1 only for the non-load-bearing self-citations that frame the program.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 2 invented entities

Load-bearing content is standard L2 geometry plus regular parametric likelihood calculus (Bahadur's setting). No fitted parameters. The one genuine addition is the generalized-estimator object and its information functional Λ; optimality then follows from Cauchy–Schwarz without extra postulates. Smoothness and common support are domain assumptions inherited from Bahadur's smooth chapters, not hidden.

axioms (4)
  • domain assumption Family {P_θ} mutually absolutely continuous with common support; Θ a connected open interval.
    Standing Bahadur assumptions from Lecture 5/11 onward; used in Lemma 1 (common support) and throughout likelihood-ratio geometry.
  • domain assumption Densities positive and continuously differentiable in θ; scores dominated in L2 near each θ so differentiation under the integral is valid (Lecture 16 Conditions 1–3).
    Required for score, identity (7), Definition 2(ii), and Λ(τ); paper states nonsmooth case is out of scope.
  • standard math L2(P_θ) Hilbert space structure: inner products, orthogonal projections, Cauchy–Schwarz.
    Bahadur Lectures 1–4 background used for W_θ, LMVUE projection, and Theorem 1.
  • ad hoc to paper A generalized estimator is assessed by Λ(τ)=(E τ')²/Var(τ), i.e. squared average slope of the standardized curve.
    Definition 3; the choice of criterion is the paper's modeling decision that makes uniform optimality available. Not forced by classical risk.
invented entities (2)
  • Generalized estimator τ: S×Θ→R with mean-zero sections, smooth in θ, positive variance independent evidence
    purpose: Replace point estimators so each sample yields a continuum of test statistics on Θ
    Definition 2; central object. Independent handle is that roots and intervals {θ:|τ̄_s(θ)|≤z} are readable from data, including at boundaries where ML fails.
  • Information utilized Λ(τ) and efficiency Eff(τ)=Λ(τ)/I independent evidence
    purpose: Family-aware assessment criterion admitting uniform optimum (the score)
    Definitions 3–4; rearranges classical information inequality as like-with-like comparison. Falsifiable in the sense that Eff can be computed for concrete models and compared to known information loss (Efron curvature).

reviewed 2026-07-31 · how reviews work

0 comments
Cite this review

Pith. "Pith review of A New Look at the Classical Estimation Problem." pith.science (2026). https://pith.science/paper/2HU2SSY2

@misc{pith2026260724890,
  author       = {Pith},
  title        = {Pith review of: A New Look at the Classical Estimation Problem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2HU2SSY2}},
  note         = {Machine review of arXiv:2607.24890}
}
Share X Bluesky LinkedIn Reddit HN
abstract

Bahadur's \emph{Lectures on the Theory of Estimation} develop the classical theory of point estimation inside the geometry of Hilbert space, and they record with unusual honesty where the theory strains: the locally best unbiased estimate depends on the parameter, a two-point parameter space yields an estimate Bahadur calls absurd, the odds ratio in binomial sampling has no unbiased estimate, and the virtues of maximum likelihood enter as heuristics and remain heuristics. We present a subset of the lectures, in Bahadur's notation and development, and at each strain make one small modification: for each value in the sample space, an estimate $\tau$ becomes a function on the parameter space rather than a point in it, the continuum of null hypotheses that Fisher described in 1955. Bahadur's own definition of an estimate, square-integrable at every distribution in the family, already supplies the domain. The payoffs are tracked lecture by lecture: estimators that exist at boundary samples where point estimates do not; an elementary lemma showing that no pointwise criterion admits a uniformly optimal estimator, which explains why admissibility, minimaxity, Bayes averaging, and unbiasedness arose as responses; assessment by information, $\Lambda(\tau)$, with the score attaining the Fisher information bound uniformly by a three-line argument; Cram\'{e}r--Rao attainment and sufficiency recovered as equality cases of that bound under two maps from point estimators to generalized estimators; and the maximum likelihood heuristics converted into exact statements about the score. Nothing classical is overturned; the classical apparatus is explained using Fisher's characterization of estimation as a continuum of significance tests.

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

141 extracted references · 14 canonical work pages · 1 internal anchor

  1. [1]

    2026 , eprint=

    Strong Likelihood Principle: Strengthening a Principle or Misunderstanding the Likelihood Function , author=. 2026 , eprint=

    2026

  2. [2]

    , title =

    Vos, Paul W. , title =

  3. [3]

    Bahadur, R. R. , editor =. R. R. Bahadur’s lectures on the theory of estimation , url =. doi:10.1214/lnms/1215091005 , publisher =

    arXiv

  4. [4]

    Bahadur, Raghu Raj , editor =. R. R. Bahadur's Lectures on the Theory of Estimation , series =. 2002 , doi =

    2002

  5. [5]

    Lehmann, E. L. and Casella, George , title =

  6. [6]

    , title =

    Vos, Paul W. , title =. Entropy , year =

  7. [7]

    Bartlett, M. S. , title =. Biometrika , year =

  8. [8]

    An infinite-dimensional geometric structure on the space of all the probability measures equivalent to a given one

    Pistone, Giovanni and Sempi, Carlo. An infinite-dimensional geometric structure on the space of all the probability measures equivalent to a given one. Ann. Stat

  9. [9]

    Estimating equations in the presence of a nuisance parameter

    Godambe, V P and Thompson, M E. Estimating equations in the presence of a nuisance parameter. Ann. Stat

  10. [10]

    Estimation in the presence of infinitely many nuisance parameters--geometry of estimating functions

    Amari, Shun-Ichi and Kumon, Masayuki. Estimation in the presence of infinitely many nuisance parameters--geometry of estimating functions. Ann. Stat

  11. [11]

    Geometrization of statistical theory , pages=

    Dual connections on the Hilbert bundles of statistical models , author=. Geometrization of statistical theory , pages=. 1987 , publisher=

    1987

  12. [12]

    Affine statistical bundle modeled on a Gaussian Orlicz--Sobolev space

    Pistone, Giovanni. Affine statistical bundle modeled on a Gaussian Orlicz--Sobolev space. Inf. Geom

  13. [13]

    Biostatistics , year =

    Confidence intervals that match Fisher's exact or Blaker's exact tests , author =. Biostatistics , year =

  14. [14]

    R Journal , year =

    Two-sided Exact Tests and Matching Confidence Intervals for Discrete Data , author =. R Journal , year =

  15. [15]

    Biometrics , year =

    Combining one sample confidence procedures for inference in the two sample case , author =. Biometrics , year =

  16. [16]

    Statistics in Medicine , year =

    A boundary-optimized rejection region test for the two-sample binomial problem , author =. Statistics in Medicine , year =

  17. [17]

    Statistics in Medicine , year =

    Confidence Intervals for Difference in Proportions for Matched Pairs Compatible with Exact McNemars or Sign Tests , author =. Statistics in Medicine , year =

  18. [18]

    2025 , url =

    R: A Language and Environment for Statistical Computing , author =. 2025 , url =

    2025

  19. [19]

    Chapter 4: Statistical Manifolds

    Lauritzen, Steffen L. Chapter 4: Statistical Manifolds. Institute of Mathematical Statistics Lecture Notes - Monograph Series

  20. [20]

    The geometry of exponential families

    Efron, Bradley. The geometry of exponential families. Ann. Stat

  21. [21]

    Mendenhall and Rodney R

    William M. Mendenhall and Rodney R. Million and Daniel E. Sharkey and Nicholas J. Cassisi , keywords =. Stage T3 squamous cell carcinoma of the glottic larynx treated with surgery and/or radiation therapy , journal =. 1984 , issn =. doi:https://doi.org/10.1016/0360-3016(84)90054-3 , url =

    work page doi:10.1016/0360-3016(84)90054-3 1984

  22. [22]

    Prednisolone in the treatment of severe malignant hypercalcaemia in metastatic breast cancer: a randomized study

    Kristensen, B and Ejlertsen, B and Holmegaard, S N and Krarup-Hansen, A and Transb l, I and Mouridsen, H. Prednisolone in the treatment of severe malignant hypercalcaemia in metastatic breast cancer: a randomized study. J. Intern. Med

  23. [23]

    Journal of the Royal Statistical Society: Series B (Methodological) , volume =

    Tibshirani, Robert , title =. Journal of the Royal Statistical Society: Series B (Methodological) , volume =. 2018 , month =. doi:10.1111/j.2517-6161.1996.tb02080.x , url =

    arXiv 2018

  24. [24]

    Stein's Estimation Rule and Its Competitors--An Empirical Bayes Approach , urldate =

    Bradley Efron and Carl Morris , journal =. Stein's Estimation Rule and Its Competitors--An Empirical Bayes Approach , urldate =

  25. [26]

    Defining the curvature of a statistical problem (with applications to second order efficiency)

    Efron, Bradley. Defining the curvature of a statistical problem (with applications to second order efficiency). Ann. Stat

  26. [27]

    The design of experiments

    Fisher, Sir Ronald Aylmer. The design of experiments

  27. [28]

    D. R. Cox , journal =. Some Problems Connected with Statistical Inference , urldate =

  28. [29]

    Objectivity and Conditionality in Frequentist Inference

    David Cox and Debra Mayo. Objectivity and Conditionality in Frequentist Inference. Error and Inference: Recent Exchanges on Experimental Reasoning, Reliability and the Objectivity and Rationality of Science

  29. [30]

    Cordeiro and Peter McCullagh , journal =

    Gauss M. Cordeiro and Peter McCullagh , journal =. Bias Correction in Generalized Linear Models , urldate =

  30. [31]

    Information and Accuracy Attainable in Estimation of Statistical Parameters , journal =

    CR Rao , year=. Information and Accuracy Attainable in Estimation of Statistical Parameters , journal =

  31. [32]

    On the mathematical foundations of theoretical statistics

    Fisher, R A. On the mathematical foundations of theoretical statistics. Philos. Trans. R. Soc. Lond

  32. [33]

    2023 , note =

    meta.shrinkage: Meta-Analyses for Simultaneously Estimating Individual Means , author =. 2023 , note =

    2023

  33. [34]

    Role of chemotherapy for advanced/recurrent gastric cancer: an individual-patient-data meta-analysis

    GASTRIC (Global Advanced/Adjuvant Stomach Tumor Research International Collaboration) Group and Oba, Koji and Paoletti, Xavier and Bang, Yung-Jue and Bleiberg, Harry and Burzykowski, Tomasz and Fuse, Nozomu and Michiels, Stefan and Morita, Satoshi and Ohashi, Yasuo and Pignon, Jean-Pierre and Rougier, Philippe and Sakamoto, Junichi and Sargent, Daniel and...

  34. [35]

    Statistical methods for research workers

    Fisher, R A. Statistical methods for research workers

  35. [36]

    M. S. Bartlett , journal =. R

  36. [37]

    Bradley Efron , journal =. R

  37. [38]

    Savage , title =

    Leonard J. Savage , title =. The Annals of Statistics , number =. 1976 , doi =

    1976

  38. [39]

    R. R. Bahadur , journal =. On the Asymptotic Efficiency of Tests and Estimates , urldate =

  39. [40]

    Approximation theorems of mathematical statistics

    Serfling, Robert J. Approximation theorems of mathematical statistics

  40. [41]

    SAMSON , journal =

    EDWARD W. SAMSON , journal =. FUNDAMENTAL NATURAL CONCEPTS OF INFORMATION THEORY , urldate =

  41. [42]

    Introduction to mathematical statistics

    Hogg, Robert V and Craig, Allen T. Introduction to mathematical statistics

  42. [43]

    Statistical Inference

    Casella, George and Berger, Roger. Statistical Inference

  43. [45]

    Encyclopedia of statistical sciences , author =

    Frequency interpretation in probability and statitical inference , volume =. Encyclopedia of statistical sciences , author =

  44. [46]

    Wasserstein and Allen L

    Ronald L. Wasserstein and Allen L. Schirm and Nicole A. Lazar , title =. The American Statistician , volume =. 2019 , publisher =. doi:10.1080/00031305.2019.1583913 , URL =

    Pith/arXiv arXiv 2019

  45. [47]

    Random variables Aren't random

    Vos, Paul W. Random variables Aren't random. Mathematics. doi:10.3390/math13050775

    work page doi:10.3390/math13050775

  46. [48]

    Semantic and cognitive tools to aid statistical science: replace confidence and significance by compatibility and surprise

    Rafi, Zad and Greenland, Sander. Semantic and cognitive tools to aid statistical science: replace confidence and significance by compatibility and surprise. BMC Med. Res. Methodol

  47. [49]

    What is mathematics?

    Courant, R and Robbins, Herbert. What is mathematics?

  48. [50]

    Principles of statistical inference

    Cox, D R. Principles of statistical inference

  49. [51]

    Modern probability theory and its applications

    Parzen, Emanuel. Modern probability theory and its applications

  50. [52]

    Some aspects of the theory of estimating equations

    Godambe, V P and Thompson, M E. Some aspects of the theory of estimating equations. J. Stat. Plan. Inference

  51. [53]

    and Lazar, Nicole A

    Wasserstein, Ronald L. and Lazar, Nicole A. , year =. The ASA Statement on p-Values: Context, Process, and Purpose , volume =. The American Statistician , publisher =. doi:10.1080/00031305.2016.1154108 , number =

    Pith/arXiv arXiv 2016

  52. [54]

    Introduction to the Practice of Statistics

    Moore, David, McCabe, George, and Craig, Bruce. Introduction to the Practice of Statistics. 2009

    2009

  53. [55]

    Journal of the Operations Research Society of Japan , author =

    Scientific thought and the refinement of human reasoning , volume =. Journal of the Operations Research Society of Japan , author =. 1960 , pages=

    1960

  54. [56]

    Scientific thought and the refinement of human reasoning

    Sir Ronald Aylmer Fisher. Scientific thought and the refinement of human reasoning. 1960

    1960

  55. [57]

    Why probability probably doesn't exist (but it is useful to act like it does) , volume =

    Spiegelhalter, David , year =. Why probability probably doesn't exist (but it is useful to act like it does) , volume =. Nature , publisher =. doi:10.1038/d41586-024-04096-5 , number =

    doi:10.1038/d41586-024-04096-5

  56. [58]

    Biased versus unbiased estimation , journal =

    Bradley Efron , abstract =. Biased versus unbiased estimation , journal =. 1975 , issn =. doi:https://doi.org/10.1016/0001-8708(75)90114-0 , url =

    work page doi:10.1016/0001-8708(75)90114-0 1975

  57. [59]

    Preprint, https://arxiv.org/abs/2412.08475 , url=

    Paul Vos , year=. Preprint, https://arxiv.org/abs/2412.08475 , url=

    Pith/arXiv arXiv

  58. [60]

    Rethinking Mean Square Error: Information, Generalized Estimation, and the James-Stein Paradox

    Vos, Paul , keywords =. Rethinking Mean Square Error: Why Information is a Superior Assessment of Estimators , publisher =. 2024 , copyright =. doi:10.48550/ARXIV.2412.08475 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2412.08475 2024

  59. [61]

    An extension of quasi-likelihood estimation

    Godambe, V P and Thompson, M E. An extension of quasi-likelihood estimation. J. Stat. Plan. Inference

  60. [62]

    Godambe and M

    Comment on `An extension of quasi-likelihood estimation' by V.P. Godambe and M. E. Thompson , year =. Journal of Statistical Planning and Inference , publisher =. doi:10.1016/0378-3758(89)90107-9 , author =

    work page doi:10.1016/0378-3758(89)90107-9

  61. [63]

    V. P. Godambe , journal =. [Inference Based on Estimating Functions in the Presence of Nuisance Parameters]: Comment , urldate =

  62. [64]

    Zeger , journal =

    Kung-Yee Liang and Scott L. Zeger , journal =. Inference Based on Estimating Functions in the Presence of Nuisance Parameters , urldate =

  63. [65]

    The Geometry of Estimating Functions in the Presence of Nuisance Parameters 1 , volume =

    Chan, Schultz and Ghosh, Malay , year =. The Geometry of Estimating Functions in the Presence of Nuisance Parameters 1 , volume =. Calcutta Statistical Association Bulletin , publisher =. doi:10.1177/0008068319980301 , number =

    work page doi:10.1177/0008068319980301

  64. [66]

    2025 , pages =

    Generalized Estimation and Information , author=. 2025 , pages =. doi:10.1007/s41884-025-00164-5 , publisher =

    doi:10.1007/s41884-025-00164-5 2025

  65. [67]

    2024 , eprint=

    Generalized Estimation and Information , author=. 2024 , eprint=

    2024

  66. [68]

    Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability

    Estimation with Quadratic Loss. Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability. 1961

    1961

  67. [69]

    Breakthroughs in statistics

  68. [70]

    Machine learning and the James--Stein estimator

    Efron, Bradley. Machine learning and the James--Stein estimator. Jpn. J. Stat. Data Sci

  69. [71]

    Samworth , title =

    Richard J. Samworth , title =. Eureka , publisher =. 2012 , URL =

    2012

  70. [72]

    Stein's Paradox in Statistics

    Efron, Bradley and Morris, Carl. Stein's Paradox in Statistics. Sci. Am

  71. [73]

    D. R. Cox and N. Reid , journal =. Parameter Orthogonality and Approximate Conditional Inference , urldate =

  72. [74]

    The Annals of Statistics , number =

    Peter McCullagh , title =. The Annals of Statistics , number =. 2002 , doi =

    2002

  73. [75]

    2010 , month = sep, publisher =

    Paul Vos and Qiang Wu , title =. 2010 , month = sep, publisher =. doi:10.1007/s10463-010-0316-3 , url =

    work page doi:10.1007/s10463-010-0316-3 2010

  74. [76]

    Vos , title =

    Suzanne Hudson and Paul W. Vos , title =. 2000 , month = dec, publisher =. doi:10.2307/3315922 , url =

    work page doi:10.2307/3315922 2000

  75. [77]

    On information and ancillarity in the presence of a nuisance parameter

    Liang, Kung-Yee. On information and ancillarity in the presence of a nuisance parameter. Biometrika

  76. [78]

    Differential-geometrical methods in statistics

    Amari, Shun-Ichi. Differential-geometrical methods in statistics

  77. [79]

    2022 , month = dec, publisher =

    Paul Vos , title =. 2022 , month = dec, publisher =. doi:10.1007/s41884-022-00085-7 , url =

    doi:10.1007/s41884-022-00085-7 2022

  78. [80]

    doi:10.2307/2335894 , url =

    Ole Barndorff-Nielsen , title =. doi:10.2307/2335894 , url =

    work page doi:10.2307/2335894

  79. [81]

    D. R. Cox , title =. doi:10.1093/biomet/asw052 , url =

    work page doi:10.1093/biomet/asw052

  80. [82]

    Brown and T

    Lawrence D. Brown and T. Tony Cai and Anirban DasGupta , title =. doi:10.1214/ss/1009213286 , url =

    arXiv

Showing first 80 references.

This paper was first reviewed by grok-4.5 on July 31, 2026.