REVIEW 4 major objections 6 minor 2 references
Characterizations of Two-Points and Other Related Distributions
T0 review · 4 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read The paper proves that simple identities among random-coefficient linear forms force exactly the two-point, uniform, and squared hyperbolic secant distributions.
desk verdict A short paper with a genuinely new uniform-distribution characterization and a plausible but unproven squared-hyperbolic-secant analogue; the proof gaps are real but mostly fixable. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The argument is carried by converting a distributional identity into a functional equation for the characteristic function $f$, then proving uniqueness of its solution. The two-point theorems reduce to the equation $\tfrac12(g(t)+1)=g^2(t/2)$, whose solution $g(t)=\cos(at)$ is selected by the method of intensively monotone operators. The uniform theorem introduces the ratio $K(t)=f(t)/f(t/2)$, which obeys the same equation $K(t)=2K^2(t/2)-1$; uniqueness is obtained by making the operator $Bg=2g^2(t/2)-1$ a contraction on the space $F$ of smooth symmetric functions with $g(0)=1$, $g''(0)=-1$, using the distance $d(g_1,g_2)=\sup_t |g_1(t)-g_2(t)|/|t^3|$. The fixed point is $K(t)=\cos(t)$, and iterating $f(t)=\cos(at)f(t/2)$, $f(t)=\cos(at)\cos(at/2)f(t/4)$, and so on gives $f(t)=\sin(2at)/(2at)$, the characteristic function of a uniform variable.
What would settle it
Search numerically for a symmetric non-degenerate distribution with finite absolute third moment whose characteristic function solves $\tfrac12(f(t)+f(t/2))f^2(t/4)=f^3(t/2)$ but is not $\sin(At)/(At)$; any such solution would disprove Theorem 3.1. The search can be run by evaluating the equation on mixtures of uniform or other symmetric distributions and checking whether the two linear forms $L_1$ and $L_2$ are identically distributed.
Extended reading notes
Core claim
On the paper's own terms, the discovery is that distributional equality or independence of particular linear statistics with random coefficients completely determines the law of the underlying variable. For two independent copies, the equality $\varepsilon X_1 \stackrel{d}{=} \tfrac12(X_1+X_2)$, with $\varepsilon$ taking values $0$ and $1$ with equal probability, holds if and only if $X$ takes only two values $\pm a$ with equal probability; the same conclusion follows if $\varepsilon X_1+(1-\varepsilon)X_2$ and $\varepsilon X_1-(1-\varepsilon)X_2$ are independent. For three copies, the identity $\xi X_1+\tfrac14(X_2+X_3)\stackrel{d}{=}\tfrac12(X_1+X_2+X_3)$, with $\xi$ taking values $\tfrac12$ and $1$ with equal probability, forces $X$ to be uniform on $(-A,A)$, and rebalancing the coefficients to $\xi X_1+\tfrac12(X_2+X_3)\stackrel{d}{=}X_1+\tfrac14(X_2+X_3)$ forces the squared hyperbolic secant law with characteristic function $t/\sinh t$, up to scale. Theorem 1.2 extends the two-point characterization to $n$ independent summands with a specially chosen random coefficient $\varepsilon_n$. The proof of Theorem 4.1 is omitted, with the paper noting that it is very similar to the uniform case.
Load-bearing premise
The uniform-distribution proof depends on the claim that the set of functions used in the contraction argument has no gaps under the distance $d(g_1,g_2)=\sup_t |g_1(t)-g_2(t)|/|t^3|$; the paper states this completeness without proof, and if that claim fails the argument that $K(t)=\cos(t)$ is the only fixed point no longer follows.
Editorial extensions
If this is right
- The two-point characterizations give a distributional test: if a symmetric variable satisfies either the equality in law or the independence condition, it is exactly a two-point variable, up to scale.
- The uniform theorem characterizes uniformity with no density assumption, using only symmetry, a finite third absolute moment, and equality in distribution of two random-coefficient linear forms.
- The squared hyperbolic secant characterization extends the same random-coefficient technique to a continuous heavy-tailed law, so the method is not limited to compactly supported distributions.
- Together with the earlier hyperbolic secant characterization, these results suggest that random-coefficient linear-form identities form a unified way to single out classical distributions across discrete, compact, and unbounded cases.
Reading between the lines
- Changing the coefficients in the linear forms is a natural experiment: the ratio method $K(t)=f(t)/f(t/2)$ may characterize a whole family of laws whose characteristic functions are infinite products of cosines or hyperbolic functions, of which uniform and squared hyperbolic secant are two examples.
- The independence version in Theorem 2.1 can be read as an identifiability result for source separation: if two random-coefficient mixtures of a symmetric source are independent, the source must be two-point, so the theorem gives a testable criterion for discrete sources.
- A concrete next step is to test whether the uniform characterization survives with only a finite second moment; if it does, the essential condition is the distributional identity itself rather than the third-moment regularity used in the proof.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes characterizations of three distribution families—two-point, uniform, and squared hyperbolic secant—through identities or independences of linear forms with random coefficients. Theorem 1.1 characterizes the two-point distribution by the identity εX1 =d 1/2(X1+X2), and Theorem 1.2 gives an n-variable analogue; Theorem 2.1 characterizes the two-point distribution by independence of two linear forms, reducing to Theorem 1.1. Theorem 3.1 claims that L1 = ξX1 + 1/4(X2+X3) and L2 = 1/2(X1+X2+X3) are identically distributed if and only if X is uniform on a symmetric interval, with a proof based on a contraction argument for K(t)=f(t)/f(t/2). Theorem 4.1 makes an analogous claim for the squared hyperbolic secant distribution, but its proof is omitted.
Significance. If the characterizations hold, they add to a small literature on distributional identities with random coefficients and extend the intensively monotone operator method beyond the Gaussian and hyperbolic secant cases. The paper's clean reduction in Theorem 2.1 and the product formula in Section 3 are attractive. However, the presentation is not self-contained: the main tool is a citation to the author's book [2], Theorem 4.1 is unproved, and the proof of Theorem 3.1 contains a false completeness assertion. The results are plausible, but as written the paper does not rigorously establish the claimed characterizations.
major comments (4)
- [Section 4, Theorem 4.1] Theorem 4.1 is the sole characterization of the squared hyperbolic secant distribution, yet its proof is omitted with only the sentence "The proof of Theorem 4.1 is very similar to that of the previous Theorem 3.1 and, therefore, is omitted." No functional equation, contraction argument, or uniqueness proof is given, so the iff claim is unverified. At minimum the authors must provide the characteristic-function equation analogous to (3.1), the definition of the appropriate K(t), and the uniqueness argument; a reader should be able to check that the identity forces f(t)=t/sinh(t) up to scale.
- [Section 3, after Eq. (3.3)] The assertion that F is a complete metric space under d(g1,g2)=sup |g1(t)-g2(t)|/|t^3| is false. A Cauchy sequence of C^3 symmetric functions in this metric need not converge to a C^3 limit (it may converge to a function with only C^1 or worse regularity), so the contraction mapping theorem cannot be applied as stated. Nevertheless, the contraction inequality d(Bg1,Bg2) ≤ 1/2 d(g1,g2) is sufficient to establish uniqueness of a fixed point in F because both K(t) and cos(t) are fixed points: if two fixed points existed, their d-distance would be at most half itself. Please replace the completeness-based argument with this direct uniqueness argument.
- [Theorem 1.2, lines after the statement] The support of εn is misstated relative to the probabilities in (1.4)-(1.5). For odd n, the probabilities assign mass to values n, n-2, ..., 1, but the text says εn takes values 0,2,...,n-1. For even n, the text says values 0,2,...,n-2 while also giving P[εn=n]; for n=3 the stated support and probabilities are already inconsistent. In addition, the proof is only the sentence that equations (1.4)-(1.5) can be solved similarly; a full reduction to Theorem 1.1 or to the intensively monotone operator method must be given. The theorem cannot be accepted in its current form.
- [Theorem 1.1 proof] The proof of Theorem 1.1 is a pointer to [2]: after defining the operator A, it says that it is easy to check that A is intensively monotone and that the family {cos(at)} is strongly E+-positive, and then appeals to Theorem 1.1.1 of [2]. Since this theorem is the foundation for Theorem 2.1 and the later sections, the authors should either state and verify the hypotheses of [2, Theorem 1.1.1] explicitly or give an elementary proof of (1.2) (e.g., by iterating the functional equation for characteristic functions).
minor comments (6)
- [Abstract] The abstract contains a typo: 'rel ated' should read 'related'.
- [Section 2] 'each with probabilities tf rac12' is a LaTeX artifact; it should read 'each with probability 1/2'.
- [Section 3, before Eq. (3.1)] The characteristic-function computation leading to (3.1) is not shown; adding one line would help the reader verify the identity.
- [Section 3, after Eq. (3.1)] The solution is written as 'g(t)=sin(At)/(At)', but the characteristic function is elsewhere denoted f; please use a consistent symbol.
- [Section 2, display equation] The joint characteristic function is written with an unmatched closing parenthesis; the expression should be E[exp{i(s+t)εX1 + i(s-t)(1-ε)X2}] = 1/2(f(s+t)+f(s-t)).
- [Section 3, definition of F] The condition |g(t)|≤1 is not justified. For a general characteristic function f with finite third moment, why is |K(t)|≤1 on [-T,T]? This follows from continuity near 0 for sufficiently small T, but the authors should state it.
Circularity Check
No construction-level circularity: the self-cited intensive monotone operator method is load-bearing but independent of the target characterizations, and the main issues are a false completeness aside and an omitted proof rather than circular reductions.
full rationale
No circular step can be exhibited with a concrete reduction. The uniqueness proofs of Theorems 1.1 and 1.2 are delegated to the author's earlier monograph [2] (Kakosyan, Klebanov and Melamed), but the cited intensively monotone operator theorem is a general uniqueness criterion whose stated hypotheses do not include the two-point or cosine conclusion; the target distributions are not assumed in the input. Thus the self-citation is load-bearing but not circular. Theorem 2.1 reduces to Theorem 1.1 via equation (2.1), which at s=t becomes equation (1.2); this is a valid reduction to an independently supplied theorem, not a circularity. Theorem 3.1 solves K(t)=2K^2(t/2)-1 by a contraction argument; the paper's claim that the metric space F is complete is asserted without proof and is in fact false, but the contraction inequality itself already gives at most one fixed point, and since both K and cos(t) are fixed points, uniqueness does not require completeness. This is a correctness gap in the text, not an equivalence of inputs and outputs by construction. Theorem 4.1 is stated with its proof omitted ('The proof of Theorem 4.1 is very similar to that of the previous Theorem 3.1 and, therefore, is omitted'); the missing proof is a verification gap for a central claim, but it is not a definitional, fitted-input, or self-citation circularity. The score of 2 reflects the presence of load-bearing self-citation and proof omissions, not a discovered circular reduction.
Assumptions & free parameters
assumptions (3)
- domain assumption The method of intensively monotone operators, including Theorem 1.1.1 and Example 1.3.1 from [2], is correct and applicable.
- ad hoc to paper The metric space F with distance d(g1,g2)=sup |g1(t)-g2(t)|/|t^3| is complete.
- ad hoc to paper The proof of Theorem 4.1 follows from the same argument as Theorem 3.1.
Cite this review
Pith. "Pith review of Characterizations of Two-Points and Other Related Distributions." pith.science (2026). https://pith.science/paper/SP274EMQ
@misc{pith2026190801865,
author = {Pith},
title = {Pith review of: Characterizations of Two-Points and Other Related Distributions},
year = {2026},
howpublished = {\url{https://pith.science/paper/SP274EMQ}},
note = {Machine review of arXiv:1908.01865}
}
read the original abstract
We provide new characterizations of two-points and some related distributions. We use properties of independence and/or identity of the distributions of suitable linear forms of random variables. Keywords: characterization of a distribution; two-points distribution; uniform distribution; squared hyperbolic secant distribution; independent or identically distributed linear forms
Reference graph
Works this paper leans on
-
[2]
A.V. Kakosyan, L.B. Klebanov and I.A. Melamed (1984) Characteriza- tion of Distributions by the Method of Intensively Monotone Operators, Lecture Notes in Mathematics, 1088, Springer-Verlag, Berlin, Heid el- berg, New York, Toronto. 7
work page 1984
-
[1]
Linear Statistics with Random Coefficients and Characterization of Hyperbolic Secant Distribution
Lev B. Klebanov (2019) Linear Statistics with Random Coeffi- cients and Characterization of Hyperbolic Secant Distribution, arX iv: 1905.09910v1, 1–6
work page Pith review arXiv 2019
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.