Pith. sign in

REVIEW 3 major objections 6 minor 58 references

A single observation and a guessed center yield valid confidence intervals for a location parameter, and the same idea improves Student-t intervals when n is small.

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

2026-07-31 03:44 UTC pith:53QPD3JT

load-bearing objection Clean Bayesian reconstructions of the classical n=1 intervals, plus a usable augmented-t extension; the n≥2 coverage claim is numerically calibrated and only lightly cross-checked. the 3 major comments →

arxiv 2607.25007 v1 pith:53QPD3JT submitted 2026-07-27 math.ST stat.MEstat.TH

Constructing and extending n = 1 Bayesian confidence intervals for location parameters in location-scale families

David Gerard This is my paper
classification math.ST stat.MEstat.TH MSC 62F2562C1062F15
keywords n=1 confidence intervalslocation-scale familiesprobability matching priorsBayes factorsaugmented t-intervalsconfidence distributionshyperbolic excess velocity
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.

Even with one draw from a normal or any continuous symmetric location-scale family whose mean and scale are both unknown, non-trivial frequentist confidence intervals for the mean exist. The paper shows those classical intervals arise from two Bayesian constructions. First, a scale-sign-invariant prior plus a point-mass prior on the standardized distance from a pre-chosen center produces equal-tailed credible intervals whose frequentist error is only O(alpha squared) larger than nominal when alpha is small. Second, the same intervals are recovered by treating the guessed center as an extra data point and inverting a Bayes-factor test monotone in the augmented t-statistic. For n at least 2 the credible-interval route loses coverage, but the augmented-t intervals stay valid for every n and have smaller expected squared half-width than Student-t whenever the true mean lies near the guess. The construction is shown on three interstellar objects hyperbolic excess velocities.

Core claim

For any continuous symmetric location-scale density differentiable at the maximizer of nu rho(nu), the equal-tailed credible interval from the prior pi(beta,nu)=|beta|^{-1} delta_nu-tilde(nu) has frequentist error alpha+O(alpha squared) as alpha goes to 0. The same n=1 intervals arise by inverting a Bayes-factor test on an augmented t-statistic; that augmentation yields valid shorter-than-Student-t intervals for every n>=2 near the guessed center.

What carries the argument

The augmented t-statistic formed by adjoining a fixed guess A to the sample, plus the right-Haar prior on the location deviation and a point mass at the maximizer of nu rho(nu). Monotonicity of the Bayes factor in |t| turns the construction into an inverted test whose critical value is calibrated for exact coverage.

Load-bearing premise

Except for the Cauchy family, the credible-interval claim is only asymptotic for high confidence levels, so the O(alpha squared) remainder must be small and the density must be differentiable at the maximizing point.

What would settle it

Simulate many single draws from a normal, form the proposed 95 percent credible intervals, and check whether empirical coverage error stays within a few parts in 10^4 of 0.05 across a grid of standardized distances from A; systematic excess beyond O(alpha squared) would refute the claim.

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

If this is right

  • With one observation an analyst can still report a valid finite-width interval for a normal mean once a plausible center A is supplied.
  • For n=2 or 3 the augmented-t intervals are substantially narrower than Student-t when |mu-A|/sigma is moderate.
  • The same construction supplies exact confidence intervals for the Cauchy location at every level at least 50 percent.
  • The method extends to any continuous symmetric location-scale family meeting the mild differentiability condition.

Where Pith is reading between the lines

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

  • The same Haar-plus-point-mass logic may generate approximate matching priors for other group-invariant problems with a single observation.
  • Because the width gain vanishes only like 1/sqrt(n), the augmented intervals stay competitive for moderate n when the guess is accurate to within roughly sqrt(2) standard deviations.
  • Replacing the single guess A by a small cloud of centers would give a natural robustification whose coverage could still be calibrated by the same double-integral argument.

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 / 6 minor

Summary. The manuscript revisits the classical problem of constructing a valid confidence interval for a normal mean from a single observation with unknown variance, i.e., intervals of the form (X+A)/2 ± η|X−A|. Rather than presenting these intervals fully formed, the author derives them from two Bayesian constructions. First (§2.1–2.3), for any continuous symmetric location-scale family differentiable at ν̃ = argmax νρ(ν), the prior π(β,ν) = |β|⁻¹ δ_ν̃(ν) yields equal-tailed credible intervals whose frequentist error is α + O(α²) as α → 0 (Theorems 1–3), exact for the Cauchy family when α ≤ 1/2 (Corollary 2). Second (§2.4), the n = 1 intervals are shown to be augmented t-intervals (treating the anchor A as an extra observation) and to coincide with inverted frequentist tests using a Bayes factor as test statistic, with a new monotonicity result for the Bayes factor under symmetric priors on the standardized effect (Theorems 6–7). For n ≥ 2 the credible-interval construction fails to maintain coverage (§3.1, Figure 2), but the augmented-t intervals remain valid, with multipliers calibrated by numerically maximizing a coverage-failure probability α(ν,η) over ν (Eqs. (39)–(41), Table 3), and have strictly smaller expected squared half-width than Student-t in a region of the parameter space (Theorems 8–10, Figure 3). An application to n = 3 hyperbolic excess velocities of interstellar objects illustrates the method.

Significance. If the results hold, the paper makes a modest but genuine contribution. It demystifies a known curiosity (n = 1 confidence intervals for a normal mean with unknown variance) by deriving it from two transparent principles rather than pulling the interval from thin air. The point-mass prior at ν̃ = argmax νρ(ν) is, to my knowledge, a new probability-matching-type prior — matching asymptotically in the confidence level rather than in n — and Theorem 3 covers general symmetric location-scale families with explicit remainders and an honest counterexample (Note 1, uniform density). The exact Cauchy result (Corollary 2) and the monotonicity-of-Bayes-factor theorem (Theorem 7) appear novel. The n ≥ 2 augmented-t intervals come with an explicit expected-width comparison (Theorem 10) and a clean honesty about where the credible-interval approach fails (§3.1, Figure 2). The n = 1 coverage claims are independently cross-validated against Portnoy's exact formula to within 10⁻⁶ (Table 1), and the author ships a public R package and a reproduction repository. The paper also corrects a typo in Portnoy's Theorem 1.1. The n ≥ 2 validity calibration is the weakest-verified link; see major comments.

major comments (3)
  1. [§3.2, Eqs. (39)-(41), Table 3] §3.2, Eqs. (39)–(41) and Table 3: the claim that the augmented-t intervals are valid for every n ≥ 2 (coverage ≥ 1−α uniformly over ν = (μ−A)/σ) is operationalized entirely by numerically maximizing α(ν,η) — a nested χ²/normal integral — over ν ∈ (0,∞), then inverting. The manuscript reports no grid or truncation details for the ν-maximization, no tolerance for the inner integral, and no independent Monte Carlo confirmation that the resulting η values in Table 3 actually deliver ≥95% coverage at the calibrated worst-case ν. This contrasts with the n = 1 case, where the worst-case ν has closed form τ(η) (S7) and Table 1 cross-checks coverage against Portnoy's exact formula (22) to within 6×10⁻⁷. Since the width-improvement claim (Theorem 10, Figure 3) is advertised precisely in the region of moderate ν where the maximum of α(ν,η) is plausibly attained, an inaccurate maximum would under-co
  2. [Theorem 9, Appendix S2] Theorem 9 and its proof (S74)–(S80): the expansion η_{α,n} = t_{1−α/2,n−1} + O(1/√n) is derived by expanding the pivot pointwise in the data, but η_{α,n} is defined through the worst case sup_ν α(ν,η), and the O_p(1/√n) terms in (S74)–(S80) depend on ν through the (X̄−A)² contribution to σ̂². Nothing in the argument controls the remainder uniformly in ν, and the maximizing ν in (39) may itself drift with n. As written, the proof establishes the rate for fixed ν at best. Either restrict the claim to fixed ν (with the constant possibly depending on ν) or provide a uniformity argument; alternatively, label the result as a heuristic supported by the Table 3 numerics, which do show the multipliers converging to within two decimals by n = 30.
  3. [Theorem 3 / Table 1] Theorem 3 statement: the theorem asserts error probability α + O(α²) without stating the domain of α for which the underlying confidence interval (2) exists (η ≥ 1/2, i.e., α ≤ 1/2, per Appendix S1). More substantively, Table 1 shows that for moderate α the credible interval can slightly under-cover: at nominal α = 0.30 the worst-case error is 0.3168 > 0.30, and at α = 0.20 it is 0.2009. The asymptotic claim is consistent with this, but readers should be told explicitly that the credible intervals are not guaranteed conservative away from the α → 0 limit, and that exact validity holds only in the Cauchy case (Corollary 2, α ≤ 1/2). A sentence in §2.1 or §2.2 would suffice.
minor comments (6)
  1. [§5] Discussion, paragraph on intervals (44)-(45): 'Waving our hands Heuristically' — stray capitalization and informal phrasing.
  2. [Figure 2] Figure 2: with 1000 replications per scenario, the Monte Carlo standard error at a true error rate of 0.05 is about 0.007; the caption mentions exact binomial intervals, which is good, but the text should note that some apparent excursions above 0.05 at ν = 1 may be within simulation error.
  3. [§2.4] Eqs. (25)-(26): the normalization of the priors 'updated with one observation of value A' would benefit from one sentence of intuition (these are posterior densities in σ² after observing A), since the notation N(A|μ₀,σ²) as a function of σ² is otherwise easy to misread.
  4. [Theorem 8 / Appendix S7] The derivation of (37) from Theorem S4 with m = 1 is correct, but Theorem 8 would read more easily if the simplification m²(n+m−1)/n → 1 were shown explicitly rather than left to the reader to verify against (S162).
  5. [References] Reference list: the arXiv date (27 Jul 2026), the JPL access dates, and 'as of 5 June 2026' are all consistent internally, but R Core Team [2026] and the v∞ values should be re-verified at proof stage; also Abramowitz and Stegun [1964] is cited only in Appendix S2 — consider noting this at first use.
  6. [Table 2] Table 2: the comparison of augmented-t thresholds to Bayes factors assumes a standard Cauchy prior on δ; this should be stated in the table caption, not only in the surrounding text.

Circularity Check

0 steps flagged

No significant circularity: Bayesian constructions recover known n=1 intervals from Haar/fat-tail priors and BF algebra, rather than assuming those intervals as inputs.

full rationale

The paper’s load-bearing claims are derived forward from standard ingredients, not by feeding the target intervals back into their own definitions. For n=1 credible intervals, the prior is fixed by right-Haar invariance on β plus a point mass at ν̃=argmax_ν νρ(ν) (an intrinsic property of the sampling density ρ); posterior tails are then expanded by Taylor (Theorem 2 / Corollary 1) and shown to match the independently computed confidence-distribution tails of the classical interval (Appendix S1), yielding Theorem 3’s α+O(α²) statement. The match is a proved equality of two separately obtained asymptotic expressions, not a definitional identity. The BF route (Theorems 6–7) likewise starts from ordinary improper/scale priors updated by the anchor A and shows monotonicity in the augmented t-statistic, so inversion recovers the same intervals as output. For n≥2 the augmented-t intervals are defined by the same pivot, with η calibrated by maximizing the explicit coverage-failure integral α(ν,η); expected-width comparisons (Theorem 10, Figure 3) follow by direct expectation under that pivot. There is no self-citation chain (sole author; classical external references), no fitted parameter re-labeled as a prediction, and no uniqueness theorem imported from the author’s prior work. Numerical calibration of η is a verification gap, not circularity. The derivation chain is therefore self-contained against its external benchmarks.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 2 invented entities

The paper rests on standard decision-theoretic invariance, ordinary Bayesian updating, and classical sampling distributions of normal mean and variance. The only modeling choices that are not forced by the group structure are the point-mass location for ν, the user-supplied anchor A, and the symmetric prior on the standardized effect size inside the Bayes factor. No new physical entities are postulated.

free parameters (3)
  • A (anchor / prior value) = application-specific (25 km/s in §4)
    User-chosen constant that centers the intervals and the prior; coverage is guaranteed for any fixed A, but expected width depends on |μ-A|/σ. In the data example A=25 km/s is taken from Engelhardt et al. (2017).
  • π(δ) prior on standardized effect size = Cauchy (default)
    Needed to evaluate numerical Bayes-factor thresholds in Table 2; Cauchy is the default choice. Monotonicity (Theorem 7) holds for any symmetric π, so the interval equivalence does not depend on the specific shape.
  • η (critical multiplier) = tabled for α=0.05, n=2..30
    Chosen by numerical maximization of the coverage-failure probability over ν so that worst-case error equals the nominal α. Not fitted to data; determined by the sampling model.
axioms (5)
  • domain assumption Sampling model belongs to a continuous symmetric location-scale family with density (1/σ)ρ((x-μ)/σ).
    Stated at the opening of §2.1; all tail expansions and invariance arguments use symmetry and the location-scale structure.
  • domain assumption ρ is differentiable at ν̃ = argmax_{ν>0} ν ρ(ν) (or at least continuous, yielding a weaker o(α) remainder).
    Required for the Taylor expansion in Theorem 2 and the quantile-matching argument of Theorem 3; Note 1 shows failure for the uniform density.
  • standard math Right Haar prior |β|^{-1} on the multiplicative group is the appropriate invariant prior when ν is treated as known.
    Invoked via the standard correspondence between equivariant decision rules and right-Haar Bayes rules (Berger 2013, §6.6), §2.1.
  • standard math Under the improper prior 1/σ^{2} shared by null and alternative, the Bayes factor is well-defined up to a common constant that cancels.
    Standard justification for improper priors on shared parameters in Bayes factors (Jeffreys, Berger–Pericchi); discussed in §5.
  • standard math Z = (X̄-μ)/(σ/√n) ~ N(0,1) independent of W^{2} = (n-1)S^{2}/σ^{2} ~ χ^{2}_{n-1}.
    Classical normal sampling facts used in Theorems 8–10 to obtain coverage and expected width of augmented-t intervals.
invented entities (2)
  • Augmented t-interval (data X1..Xn plus anchor A treated as an extra observation) independent evidence
    purpose: Provides a single construction that recovers the classical n=1 intervals and yields valid, sometimes shorter intervals for every n≥2.
    Named and analyzed as the central n≥2 method; mathematically just a t-interval on an enlarged sample, so not a new ontological object, but a distinct procedural entity in the paper.
  • Fattest-tail point-mass prior δ_ν̃ on the standardized deviation ν independent evidence
    purpose: Selects, among point-mass (and for some ρ among all proper) priors, the one that maximizes posterior tail mass and thereby matches the confidence-distribution tails.
    Derived in §2.1 from the conditional tail expansion; independent evidence is the matching with known confidence intervals and the exact Cauchy case.

pith-pipeline@v1.2.0-grok45-kimik3 · 38154 in / 3956 out tokens · 77750 ms · 2026-07-31T03:44:38.316754+00:00 · methodology

0 comments
read the original abstract

It is a surprising, modestly known fact that when given a single observation from a normal distribution with unknown mean and unknown variance, valid and non-trivial confidence intervals for the mean can be constructed. These intervals are presented in papers fully formed, providing limited intuition for how they arise or how to generalize them. We show that these intervals can be constructed in a principled way using two separate Bayesian reasonings. In the first, for any continuous symmetric location-scale family (under mild regularity conditions) with $n=1$ observation, we derive priors which produce $(1 - \alpha)100\%$ credible intervals that are, asymptotically in the confidence level $\alpha \rightarrow 0$, valid $(1 - \alpha)100\%$ confidence intervals. In the second, we show that the $n=1$ frequentist intervals can be seen as $t$-intervals augmented with a prior value, and that these augmented $t$-intervals are equivalent to inverted frequentist tests using a Bayes factor (using appropriate priors) as a test statistic. For $n \geq 2$, our credible interval approach does not maintain the confidence level. However, for $n \geq 2$, our augmented $t$-intervals produce valid confidence intervals with lower expected squared width in parts of the parameter space than the Student $t$-intervals, indicating improvements when prior knowledge is available. We demonstrate these methods on an $n = 3$ dataset of hyperbolic excess velocities of interstellar objects.

Figures

Figures reproduced from arXiv: 2607.25007 by David Gerard.

Figure 1
Figure 1. Figure 1: Posterior density of µ given X when X ∼ N(µ, σ2 ), using prior (13), and A > X. Modes are at (X + A)/2 and A + |X − A|. The areas between values |X − A| units apart are color coded. α lf lb uf ub Worst α 0.30 -0.94 -0.72 1.94 2.00 0.3168288 0.20 -1.81 -1.77 2.81 2.84 0.2009078 0.10 -4.29 -4.28 5.29 5.29 0.1000218 0.05 -9.15 -9.15 10.15 10.15 0.0500006 0.01 -47.89 -47.89 48.89 48.89 0.0100000 [PITH_FULL_IM… view at source ↗
Figure 2
Figure 2. Figure 2: Confidence error rate (y-axis, square-root scale) for 95% credible intervals based on posterior (31) for n = 2, 10, 100 (x-axis), β = 1, 5, 10 (color), and ν = 0.5, 1, 2 (facets). Exact binomial confidence intervals are plotted based on the 1000 replications for each scenario. The error rate should be at or below 0.05 (the horizontal dashed line) to be valid 95% confidence intervals. Only when ν = 1 do we … view at source ↗
Figure 3
Figure 3. Figure 3: Ratio of 95% augmented t to Student t CI expected squared half-widths (y-axis) for different values of the standardized deviation from the mean (x-axis), ν = (µ − A)/σ. Values below 1 (dashed horizontal line) indicate that the augmented t-interval has shorter expected squared width. The vertical dashed line is at √ 2, the value of ν above which the Student t-interval appears to asymptotically have shorter … view at source ↗
Figure 4
Figure 4. Figure 4: Posterior density of µ (left facet) and posterior predictive distribution for a new obser￾vation (right facet) based only on the ’Oumuamua observation. The 0.025 and 0.975 quantiles are indicated by the orange dashed lines, the value A used is indicated by the blue dotted line, and the Borisov and ATLAS observations are indicated by the green dot-dash lines. 21 [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗

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

58 extracted references · 13 canonical work pages

  1. [1]

    Abbott and J

    J. Abbott and J. Rosenblatt. Two stage estimation with one observation on the first stage. Annals of the Institute of Statistical Mathematics, 14 0 (1): 0 229--235, 1962. doi:10.1007/BF02868644

    work page doi:10.1007/bf02868644 1962

  2. [2]

    Abramowitz and I

    M. Abramowitz and I. Stegun. Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. Applied mathematics series. U.S. Government Printing Office, 1964. URL https://books.google.com/books?id=mlAs1jYzI_QC

    1964

  3. [3]

    S. Basu. Bayesian hypotheses testing using posterior density ratios. Statistics & probability letters, 30 0 (1): 0 79--86, 1996. doi:10.1016/0167-7152(95)00204-9

    work page doi:10.1016/0167-7152(95)00204-9 1996

  4. [4]

    Bayarri, D

    M. Bayarri, D. J. Benjamin, J. O. Berger, and T. M. Sellke. Rejection odds and rejection ratios: A proposal for statistical practice in testing hypotheses. Journal of Mathematical Psychology, 72: 0 90--103, 2016. doi:10.1016/j.jmp.2015.12.007

    doi:10.1016/j.jmp.2015.12.007 2016

  5. [6]

    J. Berger. Statistical Decision Theory: Foundations, Concepts, and Methods. Springer Series in Statistics. Springer New York, 2013. ISBN 9781475717273

    2013

  6. [7]

    J. O. Berger and L. R. Pericchi. Objective B ayesian methods for model selection: Introduction and comparison. In P. Lahiri, editor, Model Selection, volume 38 of Lecture Notes---Monograph Series, pages 135--193. Institute of Mathematical Statistics, 2001. ISBN 0-940600-52-8. URL https://www.jstor.org/stable/4356165

    arXiv 2001

  7. [8]

    Blachman and R

    N. Blachman and R. Machol. Confidence intervals based on one or more observations. IEEE transactions on information theory, 33 0 (3): 0 373--382, 1987. doi:10.1109/TIT.1987.1057306

    arXiv 1987

  8. [9]

    Borisov , P

    G. Borisov , P. Birtwhistle , P. Bacci , M. Maestripieri , T. Chen , D. W. E. Green , S. Nakano , H. Sato , and T. Durig . Comet C/2019 q4 (borisov) . Central Bureau Electronic Telegrams, 4666: 0 1, Sept. 2019

    2019

  9. [10]

    Casella and R

    G. Casella and R. Berger. Statistical Inference. Chapman & Hall/CRC Texts in Statistical Science. CRC Press, 2024. ISBN 9781040024010

    2024

  10. [11]

    G. S. Datta and T. J. Sweeting. Probability matching priors. In D. Dey and C. Rao, editors, Bayesian Thinking, volume 25 of Handbook of Statistics, pages 91--114. Elsevier, 2005. doi:10.1016/S0169-7161(05)25003-4

    work page doi:10.1016/s0169-7161(05)25003-4 2005

  11. [12]

    Denneau , R

    L. Denneau , R. Siverd , J. Tonry , H. Weiland , N. Erasmus , A. Fitzsimmons , J. Robinson , S. Deen , Z. T. F. Collaboration , Q.-Z. Ye , A. C. Gilmore , P. M. Kilmartin , F. Ursache , K. Korlevic , L. Poropat , M. Urbanik , E. Prosperi , M. Jaeger , G. Rhemann , H. Duin , A. Hale , Z. Moravec , C. Jacques , G. Neue , M. Alarcon , J. Licandro , P. Nichit...

    2025

  12. [13]

    T. J. DiCiccio, T. A. Kuffner, and G. A. Young. A simple analysis of the exact probability matching prior in the location-scale model. The American Statistician, 71 0 (4): 0 302--304, 2017. doi:10.1080/00031305.2016.1255662

    Pith/arXiv arXiv 2017

  13. [14]

    NIST Digital Library of Mathematical Functions

    DLMF . NIST Digital Library of Mathematical Functions . https://dlmf.nist.gov/, Release 1.2.4 of 2025-03-15. URL https://dlmf.nist.gov/. F. W. J. Olver, A. B. Olde Daalhuis , D. W. Lozier, B. I. Schneider, R. F. Boisvert, C. W. Clark, B. R. Miller, B. V. Saunders, H. S. Cohl, and M. A. McClain, eds

    2025

  14. [15]

    D. Edelman. A confidence interval for the center of an unknown unimodal distribution based on a sample of size 1. The American Statistician, 44 0 (4): 0 285--287, 1990. doi:10.1080/00031305.1990.10475740

    Pith/arXiv arXiv 1990

  15. [16]

    Engelhardt, R

    T. Engelhardt, R. Jedicke, P. Vere s , A. Fitzsimmons, L. Denneau, E. Beshore, and B. Meinke. An observational upper limit on the interstellar number density of asteroids and comets. The Astronomical Journal, 153 0 (3): 0 133, 2017. doi:10.3847/1538-3881/aa5c8a

    doi:10.3847/1538-3881/aa5c8a 2017

  16. [17]

    A. Fowlie. Neyman-- P earson lemma for B ayes factors. Communications in Statistics-Theory and Methods, 52 0 (15): 0 5379--5386, 2023. doi:10.1080/03610926.2021.2007265

    arXiv 2023

  17. [18]

    Francis and V

    G. Francis and V. Jakicic. Equivalent statistics for a one-sample t -test. Behavior Research Methods, 55 0 (1): 0 77--84, 2023. doi:10.3758/s13428-021-01775-3

    work page doi:10.3758/s13428-021-01775-3 2023

  18. [19]

    Galassi, J

    M. Galassi, J. Davies, J. Theiler, B. Gough, G. Jungman, P. Alken, M. Booth, and F. Rossi. GNU Scientific Library Reference Manual (3rd Ed.). Network Theory Limited Godalming, 2009. ISBN 0954612078. URL http://www.gnu.org/software/gsl/

    2009

  19. [20]

    Gelman, X.-L

    A. Gelman, X.-L. Meng, and H. Stern. Posterior predictive assessment of model fitness via realized discrepancies. Statistica Sinica, 6 0 (4): 0 733--760, 1996. URL https://www.jstor.org/stable/24306036

    arXiv 1996

  20. [21]

    Gelman, J

    A. Gelman, J. Carlin, H. Stern, D. Dunson, A. Vehtari, and D. Rubin. Bayesian Data Analysis, Third Edition. Chapman & Hall/CRC Texts in Statistical Science. Taylor & Francis, 2013. ISBN 9781439840955

    2013

  21. [22]

    G \"o nen, W

    M. G \"o nen, W. O. Johnson, Y. Lu, and P. H. Westfall. The B ayesian two-sample t test. The American Statistician, 59 0 (3): 0 252--257, 2005. doi:10.1198/000313005X55233

    work page doi:10.1198/000313005x55233 2005

  22. [24]

    Q. F. Gronau, A. Ly, and E.-J. Wagenmakers. Informed B ayesian t -tests. The American Statistician, 74 0 (2): 0 137--143, 2020. doi:10.1080/00031305.2018.1562983

    Pith/arXiv arXiv 2020

  23. [25]

    Guzik, M

    P. Guzik, M. Drahus, K. Rusek, W. Waniak, G. Cannizzaro, and I. Pastor-Marazuela. Initial characterization of interstellar comet 2I/Borisov . Nature Astronomy, 4 0 (1): 0 53--57, 2020. doi:10.1038/s41550-019-0931-8

    doi:10.1038/s41550-019-0931-8 2020

  24. [26]

    M. J. Hopkins, M. T. Bannister, and C. Lintott. Predicting interstellar object chemodynamics with G aia. The Astronomical Journal, 169 0 (2): 0 78, 2025. doi:10.3847/1538-3881/ad9eb3

    doi:10.3847/1538-3881/ad9eb3 2025

  25. [27]

    Jeffreys

    H. Jeffreys. The theory of probability. OUP Oxford, 1998. ISBN 9780191589676

    1998

  26. [28]

    JPL Small-Body Database: 3I/ATLAS (C/2025 N1)

    JPL . JPL Small-Body Database: 3I/ATLAS (C/2025 N1) . https://ssd.jpl.nasa.gov/tools/sbdb_lookup.html#/?sstr=3I, a . Accessed 2026-06-05; orbit solution JPL 54, solution date 2026-02-19 09:10:47

    2025

  27. [29]

    JPL Small-Body Database: 2I/Borisov (C/2019 Q4)

    JPL . JPL Small-Body Database: 2I/Borisov (C/2019 Q4) . https://ssd.jpl.nasa.gov/tools/sbdb_lookup.html#/?sstr=2I, b . Accessed 2026-06-05; orbit solution JPL 54, solution date 2024-06-24 15:49:04

    2019

  28. [30]

    JPL Small-Body Database: 1I/'Oumuamua (A/2017 U1)

    JPL . JPL Small-Body Database: 1I/'Oumuamua (A/2017 U1) . https://ssd.jpl.nasa.gov/tools/sbdb_lookup.html#/?sstr=1I, c . Accessed 2026-06-05; orbit solution JPL 16, solution date 2018-06-26 12:17:57

    2017

  29. [31]

    R. E. Kass and A. E. Raftery. Bayes factors. Journal of the American Statistical Association, 90 0 (430): 0 773--795, 1995. doi:10.1080/01621459.1995.10476572

    arXiv 1995

  30. [32]

    W. Kruskal. The monotonicity of the ratio of two noncentral t density functions. The Annals of Mathematical Statistics, 25 0 (1): 0 162--165, 1954. doi:10.1214/aoms/1177728857

    arXiv 1954

  31. [33]

    Lehmann and J

    E. Lehmann and J. Romano. Testing Statistical Hypotheses. Springer Texts in Statistics. Springer New York, 2006. ISBN 9780387276052

    2006

  32. [34]

    D. V. Lindley. Fiducial distributions and B ayes' theorem. Journal of the Royal Statistical Society. Series B (Methodological), 20 0 (1): 0 102--107, 1958. doi:10.1111/j.2517-6161.1958.tb00278.x

    arXiv 1958

  33. [35]

    McCullagh

    P. McCullagh. Conditional inference and Cauchy models. Biometrika, 79 0 (2): 0 247--259, 1992. doi:10.1093/biomet/79.2.247

    work page doi:10.1093/biomet/79.2.247 1992

  34. [36]

    K. J. Meech, R. Weryk, M. Micheli, J. T. Kleyna, O. R. Hainaut, R. Jedicke, R. J. Wainscoat, K. C. Chambers, J. V. Keane, A. Petric, L. Denneau, E. Magnier, T. Berger, M. E. Huber, H. Flewelling, C. Waters, E. Schunova-Lilly, and S. Chastel. A brief visit from a red and extremely elongated interstellar asteroid. Nature, 552 0 (7685): 0 378--381, 2017. doi...

    doi:10.1038/nature25020 2017

  35. [37]

    Mukerjee and M

    R. Mukerjee and M. Ghosh. Second-order probability matching priors. Biometrika, 84 0 (4): 0 970--975, 1997. doi:10.1093/biomet/84.4.970

    work page doi:10.1093/biomet/84.4.970 1997

  36. [38]

    H. W. Peers. On confidence points and B ayesian probability points in the case of several parameters. Journal of the Royal Statistical Society: Series B (Methodological), 27 0 (1): 0 9--16, 1965. doi:10.1111/j.2517-6161.1965.tb00581.x

    arXiv 1965

  37. [39]

    S. Portnoy. Invariance, optimality, and a 1-observation confidence interval for a normal mean. The American Statistician, 73 0 (1): 0 10--15, 2019. doi:10.1080/00031305.2017.1360796

    Pith/arXiv arXiv 2019

  38. [40]

    R: A Language and Environment for Statistical Computing

    R Core Team . R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, 2026. URL https://www.R-project.org/

    2026

  39. [41]

    C. Robert. Generalized inverse normal distributions. Statistics & Probability Letters, 11 0 (1): 0 37--41, 1991. doi:10.1016/0167-7152(91)90174-P

    work page doi:10.1016/0167-7152(91)90174-p 1991

  40. [42]

    C. P. Robert. A note on Jeffreys-Lindley paradox. Statistica Sinica, 3 0 (2): 0 601--608, 1993. URL https://www.jstor.org/stable/24304977

    arXiv 1993

  41. [43]

    C. C. Rodr \'i guez. Confidence intervals from one observation. In J. Skilling and S. Sibisi, editors, Maximum Entropy and B ayesian Methods , pages 175--182, Dordrecht, 1996. Springer Netherlands. ISBN 978-94-009-0107-0. doi:10.1007/978-94-009-0107-0\_19

    work page doi:10.1007/978-94-009-0107-0 1996

  42. [44]

    J. N. Rouder, P. L. Speckman, D. Sun, R. D. Morey, and G. Iverson. Bayesian t tests for accepting and rejecting the null hypothesis. Psychonomic bulletin & review, 16 0 (2): 0 225--237, 2009. doi:10.3758/PBR.16.2.225

    doi:10.3758/pbr.16.2.225 2009

  43. [45]

    D. B. Rubin. Bayesianly justifiable and relevant frequency calculations for the applied statistician. The Annals of Statistics, 12 0 (4): 0 1151 -- 1172, 1984. doi:10.1214/aos/1176346785

    arXiv 1984

  44. [46]

    Sans \'o , L

    B. Sans \'o , L. R. Pericchi, and E. Moreno. On the robustness of the intrinsic B ayes factor for nested models. In J. O. Berger, B. Betr \'o , E. Moreno, L. R. Pericchi, F. Ruggeri, G. Salinetti, and L. Wasserman, editors, Bayesian Robustness, volume 29 of Lecture Notes---Monograph Series, pages 155--169. Institute of Mathematical Statistics, 1996. ISBN ...

    arXiv 1996

  45. [47]

    Schweder and N

    T. Schweder and N. Hjort. Confidence, Likelihood, Probability. Cambridge Series in Statistical and Probabilistic Mathematics. Cambridge University Press, 2016. ISBN 9780521861601

    2016

  46. [48]

    D. Z. Seligman, M. Micheli, D. Farnocchia, L. Denneau, J. W. Noonan, H. H. Hsieh, T. Santana-Ros, J. Tonry, K. Auchettl, L. Conversi, M. Devogèle, L. Faggioli, A. D. Feinstein, M. Fenucci, M. Ferrais, T. Frincke, M. Gillon, O. R. Hainaut, K. Hart, A. Hoffman, C. E. Holt, W. B. Hoogendam, M. E. Huber, E. Jehin, T. Kareta, J. V. Keane, M. S. P. Kelley, T. L...

    2025

  47. [49]

    T. A. Severini, R. Mukerjee, and M. Ghosh. On an exact probability matching property of right-invariant priors. Biometrika, 89 0 (4): 0 952--957, 2002. doi:10.1093/biomet/89.4.952

    work page doi:10.1093/biomet/89.4.952 2002

  48. [50]

    Staicu and N

    A. Staicu and N. M. Reid. On probability matching priors. Canadian Journal of Statistics, 36 0 (4): 0 613--622, 2008. doi:10.1002/cjs.5550360408

    work page doi:10.1002/cjs.5550360408 2008

  49. [51]

    The probable error of a mean

    Student. The probable error of a mean. Biometrika, 6 0 (1): 0 1--25, 1908. doi:10.1093/biomet/6.1.1

    doi:10.1093/biomet/6.1.1 1908

  50. [52]

    A. G. Taylor and D. Z. Seligman. The kinematic age of 3I/ATLAS and its implications for early planet formation. The Astrophysical Journal Letters, 990 0 (1): 0 L14, 2025. doi:10.3847/2041-8213/adfa28

    doi:10.3847/2041-8213/adfa28 2025

  51. [53]

    M. M. Wall, J. Boen, and R. Tweedie. An effective confidence interval for the mean with samples of size one and two. The American Statistician, 55 0 (2): 0 102--105, 2001. doi:10.1198/000313001750358400

    work page doi:10.1198/000313001750358400 2001

  52. [54]

    Wang and G

    M. Wang and G. Liu. A simple two-sample B ayesian t -test for hypothesis testing. The American Statistician, 70 0 (2): 0 195--201, 2016. doi:10.1080/00031305.2015.1093027

    Pith/arXiv arXiv 2016

  53. [55]

    B. L. Welch and H. W. Peers. On formulae for confidence points based on integrals of weighted likelihoods. Journal of the Royal Statistical Society: Series B (Methodological), 25 0 (2): 0 318--329, 1963. doi:10.1111/j.2517-6161.1963.tb00512.x

    arXiv 1963

  54. [56]

    B. P. Welford. Note on a method for calculating corrected sums of squares and products. Technometrics, 4 0 (3): 0 419--420, 1962. doi:10.1080/00401706.1962.10490022

    arXiv 1962

  55. [57]

    G. V. Williams , H. Sato , K. Sarneczky , R. Wainscoat , D. Woodworth , and K. Meech . Minor Planets 2017 SN\_33 and 2017 U1 . Central Bureau Electronic Telegrams, 4450: 0 1, Oct. 2017

    2017

  56. [58]

    Mathematica, V ersion 14.1, 2024

    Wolfram Research , Inc. Mathematica, V ersion 14.1, 2024. URL https://www.wolfram.com/mathematica. Champaign, IL

    2024

  57. [59]

    Xie and K

    M. Xie and K. Singh. Confidence distribution, the frequentist distribution estimator of a parameter: A review. International Statistical Review, 81 0 (1): 0 3--39, 2013. doi:10.1111/insr.12000

    doi:10.1111/insr.12000 2013

  58. [60]

    Yu and P

    C. Yu and P. D. Hoff. Adaptive multigroup confidence intervals with constant coverage. Biometrika, 105 0 (2): 0 319--335, 2018. doi:10.1093/biomet/asy009

    work page doi:10.1093/biomet/asy009 2018