pith. sign in

arxiv: 2510.05739 · v3 · submitted 2025-10-07 · 🧮 math.PR · math.CO· math.ST· stat.TH

Explicit Universal Bounds for Cumulants via Moments

Jiechen Zhang This is my paper

Pith reviewed 2026-05-18 09:16 UTC · model grok-4.3

classification 🧮 math.PR math.COmath.STstat.TH
keywords cumulant boundsmoment inequalitiesdistribution-freepartition formulauniversal constantsrandom variablesindependent sums
0
0 comments X

The pith

The nth cumulant of any random variable is bounded by C_n times its nth absolute moment, with C_n growing like (n-1)! over rho to the n.

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

This paper derives explicit bounds that control the size of the nth cumulant of a random variable using only its nth absolute moment. The bounds hold for any distribution and take the form of a constant C_n times the moment, with C_n improving to roughly (n-1)! divided by 0.693 to the n power. This represents an exponential improvement compared to older bounds that grew much faster, like n to the power n. Refinements using central moments or symmetry assumptions yield even tighter constants. The result is proven using the classical relation between moments and cumulants expressed through set partitions.

Core claim

We establish explicit, universal, and distribution-free bounds for the n-th cumulant, κ_n(X), of a scalar random variable, controlled solely by an n-th order absolute moment functional M_n(X). The bounds take the form |κ_n(X)| ≤ C_n M_n(X) with C_n ∼ (n-1)!/ρ^n, which offers an exponential improvement over classical bounds where the coefficients grow superexponentially. We present a hierarchy of refinements where the rate parameter ρ increases as the functional M_n(X) incorporates more structural information, from ln 2 for raw moments to approximately 1.146 for central moments. The bounds are asymptotically efficient given the limited information of a single moment.

What carries the argument

The moment-cumulant partition formula combined with a uniform moment-product inequality that controls the sum over all set partitions.

If this is right

  • The bounds apply to every distribution with a finite nth moment, without further assumptions.
  • Switching from raw moments to central moments raises the rate rho and produces a sharper constant.
  • Symmetry assumptions allow still larger values of rho and tighter control.
  • The bounds directly control standardized cumulants of sums of independent random variables.

Where Pith is reading between the lines

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

  • The same partition argument could be adapted to bound other additive functionals that expand over set partitions.
  • Numerical checks on families with all moments finite, such as normals or exponentials, would show how close the ratio gets to the asymptotic constant.
  • Replacing the uniform product inequality with a tighter distribution-specific estimate might recover the optimal rho for each case.

Load-bearing premise

The standard moment-cumulant relation via set partitions holds and a uniform inequality bounds the relevant moment products for every random variable with a finite nth absolute moment.

What would settle it

Construct a random variable with finite nth absolute moment for which the ratio of the absolute nth cumulant to the moment exceeds the stated C_n for some large n.

read the original abstract

We establish explicit, universal, and distribution-free bounds for the $n$-th cumulant, $\kappa_n(X)$, of a scalar random variable, controlled solely by an $n$-th order absolute moment functional $M_n(X)$. The bounds take the form $\lvert\kappa_n(X)\rvert \le C_n M_n(X)$. Our principal contribution is the derivation of coefficients satisfying $C_n \sim (n-1)!/\rho^{\,n}$, which offers an exponential improvement over classical bounds where the coefficients grow superexponentially (on the order of $n^n$). We present a hierarchy of refinements where the rate parameter $\rho$ increases as the functional $M_n(X)$ incorporates more structural information. The most general bound uses the raw moment $M_n(X)=\mathsf{E}[\lvert X\rvert^n]$ with rate $\rho=\ln 2 \approx 0.693$. Using the central moment $M_n(X)=\mathsf{E}[\lvert X-\mathsf{E}[X]\rvert^n]$ improves the rate to $\rho_{\mathrm{cen}} \approx 1.146$, while assuming symmetry yields even higher rates. The proof is elementary, combining the moment-cumulant partition formula with a uniform moment-product inequality. We further prove that while these bounds are not attainable whenever the relevant coefficient is positive, they are asymptotically efficient given the limited information of a single moment. The utility of the bounds is demonstrated through an application to standardized cumulants of independent sums.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims to derive explicit, universal, distribution-free bounds |κ_n(X)| ≤ C_n M_n(X) on the nth cumulant of a scalar random variable in terms of an nth-order absolute moment functional M_n(X). The principal contribution is coefficients satisfying C_n ∼ (n-1)! / ρ^n with ρ = ln 2 ≈ 0.693 for the raw moment M_n(X) = E[|X|^n], improving to ρ_cen ≈ 1.146 for central moments and higher under symmetry. The proof combines the standard moment-cumulant partition formula with a uniform moment-product inequality derived from the power-mean inequality; the bounds are shown not to be attained but asymptotically efficient given only one moment, with an application to standardized cumulants of independent sums.

Significance. If the derivation holds, the work supplies sharper explicit constants than classical superexponential bounds (order n^n), which is useful for applications requiring control of higher cumulants from limited moment information, such as concentration or Edgeworth expansions. The elementary nature of the argument, the hierarchy of refinements based on additional structural information in M_n, and the asymptotic-efficiency claim are positive features. The distribution-free character strengthens applicability across probability theory.

major comments (2)
  1. [Proof of Theorem 1] Proof of the main bound (around the statement of Theorem 1): the uniform application of the power-mean inequality E[|X|^k] ≤ (E[|X|^n])^{k/n} to every product term arising from set partitions is asserted to be distribution-free, but the manuscript should explicitly verify that the resulting sum over Bell numbers or partitions yields precisely the claimed growth rate ρ = ln 2 rather than a looser constant; a concrete recursion or generating-function argument for the optimal ρ would strengthen the asymptotic claim.
  2. [Asymptotic efficiency discussion] Section on asymptotic efficiency: the statement that the bounds are 'asymptotically efficient given the limited information of a single moment' requires a matching lower-bound construction or sequence of distributions where |κ_n| / M_n approaches a positive fraction of C_n; without an explicit example or limit, the efficiency claim remains one-sided.
minor comments (2)
  1. [Introduction and notation] Notation for M_n(X): clarify whether the functional is always the absolute moment or switches definition between raw and central cases; a single consistent definition with subscripts would aid readability.
  2. [Application section] Application to independent sums: the standardized cumulant bound is stated but the numerical improvement over classical constants is not tabulated; adding a small comparison table for n=4,5,6 would illustrate the practical gain.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive evaluation and the constructive suggestions that help clarify the presentation. We respond to each major comment below and have incorporated revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Proof of Theorem 1] Proof of the main bound (around the statement of Theorem 1): the uniform application of the power-mean inequality E[|X|^k] ≤ (E[|X|^n])^{k/n} to every product term arising from set partitions is asserted to be distribution-free, but the manuscript should explicitly verify that the resulting sum over Bell numbers or partitions yields precisely the claimed growth rate ρ = ln 2 rather than a looser constant; a concrete recursion or generating-function argument for the optimal ρ would strengthen the asymptotic claim.

    Authors: We thank the referee for pointing this out. The growth rate ρ = ln 2 follows from applying the power-mean inequality uniformly to each product over blocks in the moment-cumulant formula and then summing over all set partitions. In the revised manuscript we have added an explicit generating-function argument: the relevant sum is bounded by the coefficients of the exponential generating function exp(e^{t/ρ} - 1) whose dominant singularity determines the precise asymptotic (n-1)! / ρ^n. We also include a short recurrence for the maximal weighted partition sum that confirms the constant is not looser than claimed. This material appears as a new remark immediately after the proof of Theorem 1. revision: yes

  2. Referee: [Asymptotic efficiency discussion] Section on asymptotic efficiency: the statement that the bounds are 'asymptotically efficient given the limited information of a single moment' requires a matching lower-bound construction or sequence of distributions where |κ_n| / M_n approaches a positive fraction of C_n; without an explicit example or limit, the efficiency claim remains one-sided.

    Authors: We agree that an explicit construction makes the efficiency statement more transparent. While the original argument already identifies the dominant partition (the one with a single large block) as the source of the leading term, we have added a concrete family of distributions in the revised version. For each n we construct a two-point random variable whose probabilities and support points are chosen so that the cumulant expansion is dominated by the same partition term; direct computation then shows that |κ_n| / M_n is asymptotically a positive multiple of (n-1)! / ρ^n. This example is placed in the section on asymptotic efficiency and demonstrates that the upper bound cannot be improved by more than a constant factor when only a single moment is available. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation uses external algebraic identities

full rationale

The paper derives the bounds |κ_n| ≤ C_n M_n by applying the standard moment-cumulant partition formula (an algebraic identity valid for any random variable with finite moments up to order n) together with the power-mean inequality E[|X|^k] ≤ (E[|X|^n])^{k/n} for k ≤ n. Both are external, distribution-free facts independent of the paper. The rate ρ is obtained by explicit summation or recursion over set partitions in the proof rather than by fitting parameters to data or by any self-referential construction. No self-citations are invoked to justify uniqueness or to forbid alternatives, and the final bound does not reduce to a renaming or redefinition of the input moment functional. The claim of asymptotic efficiency is shown by comparing the derived growth rate against the information contained in a single moment, without circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the standard combinatorial identity relating moments to cumulants and on an inequality bounding products of moments; ρ is derived rather than fitted.

axioms (2)
  • standard math Moment-cumulant partition formula
    Invoked to express κ_n as a sum over products of lower moments; standard in probability theory.
  • domain assumption Uniform moment-product inequality
    Used to bound the products appearing in the partition; introduced or applied in the proof.

pith-pipeline@v0.9.0 · 5802 in / 1294 out tokens · 29983 ms · 2026-05-18T09:16:18.050547+00:00 · methodology

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

22 extracted references · 22 canonical work pages

  1. [1]

    On large deviations.Zeitschrift für Wahrscheinlichkeitstheorie und verwandte Gebiete, 6(2):133–144, 1966

    VA Statuleviˇ cius. On large deviations.Zeitschrift für Wahrscheinlichkeitstheorie und verwandte Gebiete, 6(2):133–144, 1966

    work page 1966

  2. [2]

    The method of cumulants for the normal approxi- mation

    Hanna Döring and Peter Eichelsbacher. The method of cumulants for the normal approxi- mation. InModern Problems of Stochastic Analysis and Statistics. SP ASS 2021, pages 151–186. Springer, 2023

    work page 2021

  3. [3]

    Optimum bounds for the distributions of martingales in banach spaces.The Annals of Probability, pages 1679–1706, 1994

    Iosif Pinelis. Optimum bounds for the distributions of martingales in banach spaces.The Annals of Probability, pages 1679–1706, 1994

    work page 1994

  4. [4]

    Exact Rosenthal-type bounds.The Annals of Probability, 43(5):2511 – 2544, 2015

    Iosif Pinelis. Exact Rosenthal-type bounds.The Annals of Probability, 43(5):2511 – 2544, 2015

    work page 2015

  5. [5]

    Exact inequalities for sums of asymmetric random variables, with applications

    Iosif Pinelis. Exact inequalities for sums of asymmetric random variables, with applications. Probability Theory and Related Fields, 139(3):605–635, 2007

    work page 2007

  6. [6]

    On the bennett-hoeffding inequality

    Iosif Pinelis. On the bennett-hoeffding inequality. InAnnales de l’IHP Probabilités et statistiques, volume 50, pages 15–27, 2014

    work page 2014

  7. [7]

    Prokhorov and Yuriy A

    Yuriy V . Prokhorov and Yuriy A. Rozanov.Probability Theory: Basic Concepts, Limit Theo- rems, Random Processes, volume 157 ofDie Grundlehren der mathematischen Wissenschaften in Einzeldarstellungen. Springer-Verlag, Berlin, Heidelberg, 1969

    work page 1969

  8. [8]

    Properties and interdependence of the cumulants of a random variable.Radiophysics and Quantum Electronics, 19(8):833–839, 1976

    AA Dubkov and AN Malakhov. Properties and interdependence of the cumulants of a random variable.Radiophysics and Quantum Electronics, 19(8):833–839, 1976. 22

    work page 1976

  9. [9]

    Bounds on cumulants in terms of moments

    Iosif Pinelis. Bounds on cumulants in terms of moments. MathOverflow, Aug 2020. URL: https://mathoverflow.net/q/368761(version: 2020-08-10)

    work page 2020

  10. [10]

    Bound for cumulants of bounded random variables

    Chris Jones. Bound for cumulants of bounded random variables. Math.StackExchange, Oct

    work page

  11. [11]

    URL:https://math.stackexchange.com/q/3135549(version: 2023-10-27)

    work page arXiv 2023

  12. [12]

    V . P . Leonov and A. N. Shiryaev. On a method of calculation of semi-invariants.Theory of Probability and its Applications, 4(3):319–329, 1959

    work page 1959

  13. [13]

    Monographs on Statistics and Applied Proba- bility

    Peter McCullagh.Tensor Methods in Statistics. Monographs on Statistics and Applied Proba- bility. Chapman and Hall/CRC, 1987

    work page 1987

  14. [14]

    Multivariate distributions and the moment problem

    Christian Kleiber and Jordan Stoyanov. Multivariate distributions and the moment problem. Journal of Multivariate Analysis, 113:7–18, 2013

    work page 2013

  15. [15]

    On the moment determinacy of the distributions of compound geometric sums.Journal of Applied probability, 39(3):545–554, 2002

    Gwo Dong Lin and Jordan Stoyanov. On the moment determinacy of the distributions of compound geometric sums.Journal of Applied probability, 39(3):545–554, 2002

    work page 2002

  16. [16]

    Courier Corporation, 2013

    Jordan M Stoyanov.Counterexamples in probability. Courier Corporation, 2013

    work page 2013

  17. [17]

    Krein condition in probabilistic moment problems.Bernoulli, 6(5):939–949, 2000

    Jordan Stoyanov. Krein condition in probabilistic moment problems.Bernoulli, 6(5):939–949, 2000

    work page 2000

  18. [18]

    Stieltjes classes for moment-indeterminate probability distributions.Journal of Applied Probability, 41(A):281–294, 2004

    Jordan Stoyanov. Stieltjes classes for moment-indeterminate probability distributions.Journal of Applied Probability, 41(A):281–294, 2004

    work page 2004

  19. [19]

    Moment determinacy of powers and products of nonnegative random variables.Journal of Theoretical Probability, 28(4):1337–1353, 2015

    Gwo Dong Lin and Jordan Stoyanov. Moment determinacy of powers and products of nonnegative random variables.Journal of Theoretical Probability, 28(4):1337–1353, 2015

    work page 2015

  20. [20]

    New checkable conditions for moment determinacy of probability distributions.Theory of Probability & Its Applications, 65(3):497–509, 2020

    Jordan M Stoyanov, Gwo Dong Lin, and Peter Kopanov. New checkable conditions for moment determinacy of probability distributions.Theory of Probability & Its Applications, 65(3):497–509, 2020

    work page 2020

  21. [21]

    T. P . Speed. Cumulants and partition lattices.Australian & New Zealand Journal of Statistics, 25(2):378–388, 1983

    work page 1983

  22. [22]

    Cambridge University Press, 2009

    Philippe Flajolet and Robert Sedgewick.Analytic Combinatorics. Cambridge University Press, 2009. 23

    work page 2009