{"id":"5329c73b-9a19-4ac0-9ae6-9c00fa2f3722","arxiv_id":"2606.19726","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Derives exact finite-sample CDF and PDF for the Behrens-Fisher statistic in beta-function form by reducing the problem to harmonic measure on spherical wedges via a Laplace-Dirichlet formulation.","lead":"The paper reformulates the Behrens-Fisher problem of comparing means from two normal samples with unknown unequal variances as a Laplace equation on a sphere, yielding exact CDF and PDF expressions via beta functions that depend only on sample sizes and variance ratio. A smart generalist might read it to understand a new geometric route to exact small-sample inference without variance-equality assumptions or heavy approximations.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"The orthogonal decomposition must map the studentized mean difference exactly onto a spherical wedge whose harmonic measure at the origin is given by a beta-function expression depending only on n1, n2 and the variance ratio.","rationale":"The reader's weakest_assumption is precisely the step whose correctness is required for the beta-function claim to follow. Because the full derivation is not reproduced here, the same load-bearing point remains unverified; the special-case check above would settle it directly.","tokens_in":1715,"tokens_out":406,"duration_ms":42086,"concrete_test":"Fix n1=3, n2=3 and variance ratio \rho=1; substitute into the claimed beta-function CDF and confirm it equals the exact CDF of Student's t with 4 degrees of freedom (which is an incomplete-beta expression). Then repeat for \rho=4 and compare the resulting numerical values at three interior points against direct Monte-Carlo simulation of the Behrens-Fisher statistic (10^7 replicates); any discrepancy >0.5% at any point falsifies the reduction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that after removing the two mean directions, the remaining residual vectors (in dimensions n1-1 and n2-1) together with the normalized mean-difference constraint define a region on the product sphere whose indicator function, when solved via the Laplace-Dirichlet problem, yields a value at the origin expressible in closed beta form. This holds only if (i) the constraint surface is a constant-angle wedge (or union of wedges) whose opening depends on the variance ratio in a manner that preserves the beta integral representation, and (ii) the induced measure on the sphere coincides with the harmonic measure without additional Jacobian factors from the unequal scalings. The abstract states the reduction but supplies no explicit verification that the resulting boundary-value problem on the wedge produces precisely the claimed beta expressions rather than a more general hypergeometric or series form.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","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.","tokens_in":1915,"tokens_out":535,"duration_ms":14862,"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":[{"comment":"§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.","section":"§3.2"},{"comment":"§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.","section":"§4.1"}],"minor_comments":[{"comment":"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.","section":"§2"},{"comment":"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.","section":"§5"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"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.","responses":[{"response":"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_made":"yes","referee_comment":"[§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."},{"response":"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 \rho = 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_made":"yes","referee_comment":"[§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."}],"tokens_in":1398,"tokens_out":415,"duration_ms":10320,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","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.\n\nThis 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.\n\nThe 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.\n\nThe 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.\n\nI 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.","headline":"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.","tokens_in":2412,"tokens_out":435,"would_cite":false,"duration_ms":24759,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"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.","keywords":["Behrens-Fisher problem","Laplace equation","harmonic measure","beta function","normal samples","unequal variances","spherical wedge","finite-sample distribution"],"falsifier":"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.","tokens_in":2607,"feed_emoji":"","tokens_out":608,"duration_ms":14062,"temperature":0.7,"pith_summary":"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.","feed_headline":"Beta functions solve Behrens-Fisher distribution exactly","feed_subtitle":"Laplace-Dirichlet problem on spherical wedge yields CDF and PDF depending only on sample sizes and variance ratio.","key_machinery":"Laplace-Dirichlet boundary value problem on the spherical wedge, whose value at the origin equals the distribution function of the studentized difference.","core_discovery":"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.","pith_inferences":["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."],"forward_implications":["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."],"fun_headline_variants":["Behrens-Fisher solved with beta functions","Laplace equation for Behrens-Fisher distribution","Beta form of Behrens-Fisher from spherical wedge","Exact Behrens-Fisher CDF via Laplace-Dirichlet"],"cache_read_input_tokens":64,"weakest_assumption_plain":"The orthogonal decomposition separates mean and residual components without loss of distributional information for the studentized statistic.","fun_headline_variants_meta":{"raw":{"variants":["Behrens-Fisher solved with beta functions","Laplace equation for Behrens-Fisher distribution","Beta form of Behrens-Fisher from spherical wedge","Exact Behrens-Fisher CDF via Laplace-Dirichlet"]},"model":"grok-4.3","cost_usd":0.004817,"raw_usage":{"total_tokens":2349,"prompt_tokens":629,"num_sources_used":0,"completion_tokens":61,"cost_in_usd_ticks":48174500,"prompt_tokens_details":{"text_tokens":629,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1659,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":629,"tokens_out":61,"duration_ms":12021,"temperature":1.0,"reasoning_tokens":1659,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-26T15:36:24.980936+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"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.","supporting_citations":[],"review_version":1}