REVIEW 3 minor 21 references
On the L{\'e}vy concentration function of Gaussian quadratic forms with applications to second order U-statistics
T0 review · 0 major / 3 minor · reviewed 2026-06-25 · grok-4.3
Pith's one-line read An adaptive upper bound for the Lévy concentration function of weighted noncentral chi-square sums is derived.
desk verdict The paper gives an adaptive Levy concentration bound for weighted noncentral chi-square sums that drops the usual restrictions on the number, size, and signs of the lambda coefficients. 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 adaptive upper bound on the Lévy concentration function Q_S(ε) for the infinite sum S of weighted noncentral chi-squares.
What would settle it
Finding sequences λ_k and μ_k where the supremum probability in an interval of length ε exceeds the stated bound by more than a constant factor for arbitrarily small ε.
Extended reading notes
Core claim
We provide an upper-bound for the Lévy concentration function Q_S(ε) where S is a weighted sum of noncentral chi-square random variables S := sum λ_k (Z_k^2 - 1) + μ_k Z_k with independent standard Gaussians Z_k. Our bound is adaptive in that it recovers Gaussian type estimates if the l2 norm of λ is negligible compared to that of μ and chi-square estimates otherwise. The bound generalizes existing ones by making no assumptions on the number of nonzero |λ_k|, the size of the minimal |λ_k|, or the signs of λ_k. We apply the bound to quadratic forms arising in limit theorems for second-order U-statistics.
Load-bearing premise
That S equals the given infinite sum involving square-summable coefficient sequences and independent standard normal random variables.
Editorial extensions
If this is right
- Applies directly to limiting distributions of second-order U-statistics.
- Generalizes previous concentration bounds for such quadratic forms.
- Works for both finite and infinite sums without truncation assumptions.
- Recovers known Gaussian and chi-square concentration estimates as special cases.
Reading between the lines
- The bound could be used to obtain Berry-Esseen type rates for U-statistics.
- It might extend to vector-valued versions or other functionals in probability.
- Applications in high-dimensional data analysis where quadratic forms appear in test statistics.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper derives an explicit upper bound on the Lévy concentration function Q_S(ε) = sup_x P(x < S ≤ x + ε) for the random variable S = ∑_{k=1}^∞ λ_k (Z_k² - 1) + μ_k Z_k, where {Z_k} are i.i.d. standard normals and {λ_k}, {μ_k} are square-summable real sequences. The bound is adaptive: it recovers (up to constants) Gaussian-type estimates when ||λ||_2 is negligible relative to ||μ||_2 and chi-square-type estimates in the reverse regime. It generalizes prior results by imposing no restrictions on the number of nonzero |λ_k|, the minimal |λ_k|, or the signs of the λ_k. The bound is applied to quadratic forms appearing as limits of second-order U-statistics.
Significance. If the stated bound holds with the claimed adaptivity and generality, the result supplies a practical tool for small-ball probability estimates in the non-Gaussian limits that arise in U-statistic theory. The absence of lower bounds on min |λ_k| or cardinality restrictions on the support of λ distinguishes the work from many existing concentration inequalities for quadratic forms and could facilitate sharper analysis in settings where the relative sizes of the linear and quadratic coefficients vary.
minor comments (3)
- The abstract and introduction would benefit from a brief comparison table or explicit statement of the constants appearing in the new bound versus the constants in the Gaussian and chi-square regimes it recovers.
- Notation for the sequences λ and μ is introduced in the abstract but the precise statement of the main theorem (presumably in §3 or §4) should restate the square-summability assumption explicitly to avoid any ambiguity about the domain of the bound.
- In the applications section, the concrete U-statistic examples would be clearer if the corresponding λ and μ sequences were written out explicitly rather than left in implicit form.
Simulated Author's Rebuttal
We thank the referee for their positive assessment of the manuscript, including the recognition of its adaptivity and generality, and for the recommendation of minor revision.
Circularity Check
No significant circularity; derivation is self-contained
full rationale
The paper derives an explicit upper bound on the Lévy concentration function Q_S(ε) for the random variable S defined directly as the L²-convergent series ∑ λ_k (Z_k² - 1) + μ_k Z_k with square-summable coefficients. The bound is obtained from first-principles probabilistic estimates that interpolate between Gaussian and chi-square regimes without fitting any parameters to data, without renaming known empirical patterns, and without load-bearing self-citations that reduce the central claim to prior fitted quantities. The representation of S is the standard definition of the limiting object, not a constructed input that forces the output bound. No step in the derivation chain reduces by construction to the inputs.
Assumptions & free parameters
assumptions (2)
- standard math Z_k are independent standard Gaussian random variables
- domain assumption λ_k and μ_k are real-valued square-summable sequences
Cite this review
Pith. "Pith review of On the L{\'e}vy concentration function of Gaussian quadratic forms with applications to second order U-statistics." pith.science (2026). https://pith.science/paper/3ZVHBNZX
@misc{pith2026260625441,
author = {Pith},
title = {Pith review of: On the L\'evy concentration function of Gaussian quadratic forms with applications to second order U-statistics},
year = {2026},
howpublished = {\url{https://pith.science/paper/3ZVHBNZX}},
note = {Machine review of arXiv:2606.25441}
}
abstract
We provide an upper-bound for the L{\'e}vy concentration function: $$ Q_{S}(\varepsilon):= \sup_{x \in\mathbb{R}}\mathbb{P} (x < S \leq x+\varepsilon) $$ where $S$ is a weighted sum of noncentral chi-square random variables: $$ S:= \sum_{k=1}^\infty \lambda_k (Z_k^2 - 1) + \mu_kZ_k $$ Here, $\{Z_k\}_{k=1}^\infty$ is a sequence of independent standard Gaussian random variables and $\{\lambda_k\}_{k=1}^\infty, \{\mu_k\}_{k=1}^\infty$ are real valued, square summable sequences. Random variables of this type often appear as limiting distributions of second order U-statistics. Our bound is adaptive, in that it recovers (up to constant factors) Gaussian type concentration function estimates if $\|\lambda\|_2$ is negligible compared to $\|\mu\|_2$ and chi-square estimates if $\|\mu\|_{2}$ is negligible compared to $\|\lambda\|_2$. Our bound generalizes existing bounds in various ways. In particular, we make no assumptions regarding the number of nonzero $|\lambda_k|$ or the size of the minimal $|\lambda_k|$, nor do we make any assumptions on the signs of $\lambda_k$. Finally, we apply our bound to some examples of interest, specifically quadratic forms that arise in limit theorems for second-order U-statistics.
Reference graph
Works this paper leans on
-
[1]
2009 , publisher=
Approximation theorems of mathematical statistics , author=. 2009 , publisher=
2009
-
[2]
, TITLE =
Petrov, Valentin V. , TITLE =. 1995 , PAGES =
1995
-
[3]
Bernoulli , volume=
Large ball probabilities, Gaussian comparison and anti-concentration , author=. Bernoulli , volume=. 2019 , publisher=
2019
-
[4]
International Conference on Stochastic Methods , pages=
Two--Sided Bounds for PDF’s Maximum of a Sum of Weighted Chi-square Variables , author=. International Conference on Stochastic Methods , pages=. 2020 , organization=
2020
-
[5]
Esseen, Carl-Gustav , title =
-
[6]
and Chistyakov, Gennadiy P
Bobkov, Sergey G. and Chistyakov, Gennadiy P. , title =
-
[7]
, title=
Ferguson, Thomas S. , title=
-
[8]
and Liu, Xing and Duncan, Andrew B
Huang, Kevin H. and Liu, Xing and Duncan, Andrew B. and Gandy, Axel , title =. Proceedings of Machine Learning Research , year =
Show all 21 references
-
[9]
Theory of Probability & Its Applications , year =
Yanushkevichiene, Oleg , title =. Theory of Probability & Its Applications , year =
-
[10]
, title =
Alberink, I.B. , title =
-
[11]
and Bentkus, V
Alberink, I.B. and Bentkus, V. , title =. Theory of Probability & Its Applications , year =
-
[12]
, title=
G\"otze, F. , title=. Zeitschrift für Wahrscheinlichkeitstheorie und verwandte Gebiete , year=
-
[13]
Yurinskii, V. V. , title=. Soviet Mathematics - Doklady , year=
-
[14]
and Singh, S
Chasapis, G. and Singh, S. and Tkocz, T. , title =
-
[15]
Large ball probabilities, Gaussian comparison and anti-concentration , journal =
G. Large ball probabilities, Gaussian comparison and anti-concentration , journal =. 2019 , volume =
2019
-
[16]
arXiv preprint arXiv:2406.12437 , year=
Slow rates of approximation of U-statistics and V-statistics by quadratic forms of Gaussians , author=. arXiv preprint arXiv:2406.12437 , year=. 2406.12437 , archivePrefix=
-
[17]
Bentkus, V. and G. Optimal Bounds in Non-Gaussian Limit Theorems for U-Statistics , journal =. 1999 , volume =. doi:10.1214/aop/1022677269 , url =
1999 doi
-
[18]
The Annals of Probability , pages=
Optimal rates of convergence in the CLT for quadratic forms , author=. The Annals of Probability , pages=. 1996 , publisher=
1996
-
[19]
and Olkin, I
Marshall, A.W. and Olkin, I. and Arnold, B.C. , title =
-
[20]
Mathematical Research Letters , volume=
Carbery, Anthony and Wright, James , title =. Mathematical Research Letters , volume=. 2001 , pages=
2001
-
[21]
I. B. Alberink and V. Bentkus , title =. Acta Applicandae Mathematicae , volume =. 1999 , doi =
1999
Reviewed June 25, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.