A Laplace equation approach to the Behrens--Fisher problem

Jong Sung Kim; Nagananda K G

REVIEW 2 major objections 2 minor 3 references

A Laplace equation approach to the Behrens--Fisher problem

T0 review · 2 major / 2 minor · reviewed 2026-06-26 · grok-4.3

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

Pith's one-line read The Behrens-Fisher distribution for two normal samples with unknown unequal variances reduces to beta-function expressions via a Laplace equation on a spherical wedge.

desk verdict The paper reduces the Behrens-Fisher problem to a Laplace-Dirichlet problem on spherical wedges and claims exact beta-function forms for the CDF and PDF, but the abstract supplies no derivations to back the reduction. read the letter →

arxiv 2606.19726 v1 pith:FV2HSVZU submitted 2026-06-18 math.ST stat.TH

Nagananda K G , Jong Sung Kim This is my paper · ORCID

classification math.STstat.TH

keywords Behrens-FisherproblemLaplaceequationharmonicmeasurebetafunctionnormalsamplesunequalvariancessphericalwedgefinite-sampledistribution

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper formulates the Behrens-Fisher problem as a partial differential equation on the sphere. An orthogonal decomposition isolates the studentized difference of sample means as a scale-invariant geometric constraint whose probability equals the harmonic measure at the origin of a Laplace-Dirichlet problem. This identification produces exact finite-sample formulas for the cumulative distribution function and density that involve only beta functions of the sample sizes and the variance ratio. The resulting expressions are directly evaluable in standard software and admit a Gegenbauer series expansion plus sharp tail asymptotics.

What carries the argument

Laplace-Dirichlet boundary value problem on the spherical wedge, whose value at the origin equals the distribution function of the studentized difference.

What would settle it

Numerical evaluation of the beta-function formula for chosen sample sizes and variance ratio, compared against direct Monte Carlo sampling of the studentized difference of means from two independent normals.

Watch

Extended reading notes

Core claim

The Behrens-Fisher law admits exact finite-sample representations for its cumulative distribution function and probability density function in terms of beta functions, obtained by identifying the studentized mean difference with harmonic measure at the origin of a Laplace-Dirichlet problem on a spherical wedge whose geometry is fixed by the sample sizes and variance ratio.

Load-bearing premise

The orthogonal decomposition separates mean and residual components without loss of distributional information for the studentized statistic.

Editorial extensions

If this is right

The cumulative distribution and density become computable from elementary beta functions without simulation or numerical quadrature.
Quantiles and critical values are available in commercial spreadsheet software.
A Gegenbauer separation-of-variables series supplies an alternative explicit expansion whose coefficients are closed-form Beta-Gamma expressions.
Sharp tail expansions with explicit leading constants follow directly from the boundary-value representation.

Reading between the lines

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

The same harmonic-measure reduction may apply to generalizations with three or more samples or with elliptical rather than normal errors.
The geometric view on the sphere could link the Behrens-Fisher law to other scale-invariant statistics whose distributions are likewise harmonic measures.
Exact dependence on the variance ratio alone may simplify analytic power calculations for the associated hypothesis test.

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.

Desk Editor's Note

read the letter

The core claim is a new route to the Behrens-Fisher distribution: an orthogonal decomposition isolates the studentized mean difference as a scale-invariant constraint on the product sphere, turning the problem into a harmonic-measure evaluation whose value at the origin is asserted to be a beta function of the sample sizes and variance ratio. A Gegenbauer series for the extension is also given.

This is new in the technical sense. The literature on Behrens-Fisher has long sought closed forms, and mapping the problem to a standard boundary-value problem on wedges is not the usual path. If the algebra closes, the accessibility in Excel and similar packages would be a practical plus for small-sample work.

The soft spot is the missing verification. The abstract states the beta representations and the Gegenbauer coefficients but shows none of the steps that confirm the wedge angle induced by the variance ratio produces exactly those integrals rather than a more general hypergeometric or series result. The stress-test note correctly flags the potential Jacobian issue from unequal scalings; without an explicit check that the induced measure on the sphere matches the harmonic measure, the central identification remains unproven on the page. No numerical comparisons to known cases or simulation checks appear in the supplied material.

The paper is aimed at statisticians who need exact finite-sample inference for two normal samples with unequal variances. A reader already working on that narrow problem could extract the claimed formulas and test them directly. It is not positioned as a broad methodological advance.

I would send it to peer review. The approach is coherent on its own terms and the target is a long-standing problem, so referees can check the algebra and the boundary conditions. The current write-up is too thin on the derivations to stand alone.

Referee Report

2 major / 2 minor

Summary. The manuscript develops a PDE formulation of the Behrens-Fisher problem for two independent normal samples with unknown unequal variances. An orthogonal decomposition isolates the mean directions from the residual components, recasting the studentized mean difference as a scale-invariant geometric constraint on the product sphere. This reduces the distributional problem to computing spherical wedge probabilities via the value at the origin of a Laplace-Dirichlet problem. The paper derives exact finite-sample CDF and PDF expressions in terms of beta functions depending only on the sample sizes and variance ratio, obtains a Gegenbauer expansion for the harmonic extension, and provides sharp tail asymptotics.

Significance. If the central derivations hold, the work supplies closed-form beta-function representations for the Behrens-Fisher law that are directly evaluable in standard software, together with an explicit Gegenbauer series and tail expansions. These constitute a concrete advance in placing a classic finite-sample distribution into accessible special-function form, with potential practical value for inference procedures that rely on the exact law.

major comments (2)

[§3.2] §3.2, after Eq. (18): the reduction of the harmonic measure on the product sphere (dimensions n1-1 and n2-1) to a single beta integral requires that the metric factor arising from the unequal scalings and the wedge angle (determined by the variance ratio) cancel exactly; the manuscript states the final beta form but does not display the explicit change-of-variables Jacobian or the verification that no residual hypergeometric factor remains.
[§4.1] §4.1, Eq. (27): the claimed CDF expression must recover the central t-distribution when the variance ratio equals 1; an explicit substitution check confirming that the beta parameters collapse to the known t-CDF form (or its incomplete-beta equivalent) is needed to confirm the reduction is free of hidden constants.

minor comments (2)

[§2] Notation for the product-sphere measure should be introduced once in §2 and used consistently; the current alternation between surface measure and normalized harmonic measure is occasionally ambiguous.
[§5] The Gegenbauer coefficients in §5 are given in closed Beta-Gamma form; a short table of the first few coefficients for representative (n1,n2) pairs would aid verification.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and for identifying two points where additional explicit verification would strengthen the presentation. Both comments concern omitted intermediate steps rather than errors in the final claims; we will expand the derivations accordingly.

read point-by-point responses

Referee: [§3.2] §3.2, after Eq. (18): the reduction of the harmonic measure on the product sphere (dimensions n1-1 and n2-1) to a single beta integral requires that the metric factor arising from the unequal scalings and the wedge angle (determined by the variance ratio) cancel exactly; the manuscript states the final beta form but does not display the explicit change-of-variables Jacobian or the verification that no residual hypergeometric factor remains.

Authors: We agree that the change-of-variables step merits an explicit Jacobian computation. In the revision we will insert the coordinate transformation from the product-sphere measure to the single angular variable, display the metric factor arising from the unequal scalings, and verify that it cancels exactly against the wedge-angle term, leaving a pure beta integral with no residual hypergeometric factor. revision: yes
Referee: [§4.1] §4.1, Eq. (27): the claimed CDF expression must recover the central t-distribution when the variance ratio equals 1; an explicit substitution check confirming that the beta parameters collapse to the known t-CDF form (or its incomplete-beta equivalent) is needed to confirm the reduction is free of hidden constants.

Authors: We accept that an explicit substitution check is desirable for transparency. The revised manuscript will contain a short paragraph (or appendix entry) performing the substitution ho = 1, showing that the two beta parameters reduce to the standard incomplete-beta representation of the central t CDF with the appropriate degrees of freedom and confirming the absence of extraneous constants. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; derivation is self-contained from standard normal model

full rationale

The paper starts from the standard two-sample normal model with unknown variances, applies an orthogonal decomposition of mean and residual components, and recasts the studentized mean difference as a scale-invariant geometric constraint on the product sphere. This is identified with harmonic measure for a Laplace-Dirichlet problem whose solution at the origin yields beta-function expressions depending only on sample sizes and variance ratio. No steps reduce by construction to fitted inputs, self-definitions, or load-bearing self-citations; the Gegenbauer expansion and tail asymptotics are presented as derived consequences rather than inputs. The central claim remains independent of its own outputs.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

The central claim rests on the standard normality and independence assumptions for the two samples plus the validity of the orthogonal decomposition that produces the spherical-wedge geometry; no free parameters are introduced beyond the observable sample sizes and the unknown variance ratio that parametrizes the target distribution.

assumptions (1)

domain assumption The two samples are independent and each is drawn from a normal distribution.
This is the classical setup of the Behrens-Fisher problem invoked throughout the abstract.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Laplace equation approach to the Behrens--Fisher problem." pith.science (2026). https://pith.science/paper/FV2HSVZU

@misc{pith2026260619726,
  author       = {Pith},
  title        = {Pith review of: A Laplace equation approach to the Behrens--Fisher problem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FV2HSVZU}},
  note         = {Machine review of arXiv:2606.19726}
}

read the original abstract

We develop a partial differential equation formulation of the Behrens-Fisher problem for two independent normal samples with unknown and unequal variances. An orthogonal decomposition separates mean and residual components (corresponding to the centered within-sample variation left after removal of the mean directions) and recasts the studentized difference of sample means as a scale-invariant geometric constraint. This reduction transforms the distributional problem into the evaluation of spherical wedge probabilities, which are identified with harmonic measure and with the value at the origin of a Laplace-Dirichlet boundary value problem. From this framework, we derive exact finite-sample representations for the cumulative distribution function and the probability density function in terms of beta functions, with dependence only on the sample sizes and the variance ratio. These representations place the Behrens-Fisher law in a standard special-function form that is directly accessible in widely available commercial software -- including Microsoft Excel -- thereby facilitating distributional evaluation and quantile computation. We also obtain a Gegenbauer separation-of-variables expansion for the associated harmonic extension and its threshold derivative, with coefficients in closed Beta-Gamma form, and derive sharp tail expansions with explicit leading constants and higher-order corrections.

Figures

Figures reproduced from arXiv: 2606.19726 by the authors.

**Figure 1.** , (n1, n2) is held fixed while κ varies, so that the effect of heteroscedasticity on the exact finite-sample law can be examined directly; the cdf is shown in Fig. 1a and the pdf is shown in Fig. 1b. In Fig. 2a, κ is fixed and the sample sizes are varied, in order to display the dependence of the distribution on D = n1 +n2 −1. In Fig. 2b, (n1, n2, κ) are varied to see the tail behavior. Since the graphs are computed… view at source ↗

**Figure 2.** Effect of sample sizes (n1, n2, κ) on the density function. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 1 canonical work pages

[1]

and Gurland, J

Asiribo, O. and Gurland, J. (1989), ‘Some simple approximate solutions to the Behrens–Fisher problem’,Communications in Statistics - Theory and Methods18(4), 1201–1216. Behrens, W. U. (1929), ‘Ein beitrag zur fehlerberechnung bei wenigen beobachtungen’,Land- wirtschaftliche Jahrb¨ ucher68, 807–837. A contribution to error estimation with few observations....

1989
[2]

Chaturvedi, A., Bapat, S. R. and Joshi, N. (2019), ‘Second-order approximations for a multivariate analog of the Behrens–Fisher problem through a three-stage procedure’,Communications in Statistics - Theory and Methods49(14), 3466–3480. Chen, C., Li, Y., Liang, K. and Du, J. (2022), ‘A test for the Behrens–Fisher problem based on the method of variance es...

work page Pith review arXiv 2019
[3]

and Zhu, T

Zhang, J.-T., Zhou, B., Guo, J. and Zhu, T. (2021), ‘Two-sample Behrens-Fisher problems for high- dimensional data: A normal reference approach’,Journal of Statistical Planning and Inference 213, 142–161. Zhou, B., Guo, J. and Zhang, J.-T. (2017), ‘High-dimensional general linear hypothesis testing under heteroscedasticity’,Journal of Statistical Planning...

2021

Pith tools

Reviewed June 26, 2026 · model on record in the stance chip above.

[1] [1]

and Gurland, J

Asiribo, O. and Gurland, J. (1989), ‘Some simple approximate solutions to the Behrens–Fisher problem’,Communications in Statistics - Theory and Methods18(4), 1201–1216. Behrens, W. U. (1929), ‘Ein beitrag zur fehlerberechnung bei wenigen beobachtungen’,Land- wirtschaftliche Jahrb¨ ucher68, 807–837. A contribution to error estimation with few observations....

1989

[2] [2]

Chaturvedi, A., Bapat, S. R. and Joshi, N. (2019), ‘Second-order approximations for a multivariate analog of the Behrens–Fisher problem through a three-stage procedure’,Communications in Statistics - Theory and Methods49(14), 3466–3480. Chen, C., Li, Y., Liang, K. and Du, J. (2022), ‘A test for the Behrens–Fisher problem based on the method of variance es...

work page Pith review arXiv 2019

[3] [3]

and Zhu, T

Zhang, J.-T., Zhou, B., Guo, J. and Zhu, T. (2021), ‘Two-sample Behrens-Fisher problems for high- dimensional data: A normal reference approach’,Journal of Statistical Planning and Inference 213, 142–161. Zhou, B., Guo, J. and Zhang, J.-T. (2017), ‘High-dimensional general linear hypothesis testing under heteroscedasticity’,Journal of Statistical Planning...

2021