arxiv: 2605.11372 · v1 · submitted 2026-05-12 · 🧮 math.ST · stat.TH

Recognition: 3 theorem links

· Lean Theorem

The Geometry of Spectral Fluctuations: On Near-Optimal Conditions for Universal Gaussian CLTs, with Statistical Applications

Wang Zhou, Yanqing Yin

Pith reviewed 2026-05-13 02:05 UTC · model grok-4.3

classification 🧮 math.ST stat.TH

keywords linear spectral statisticssample covariance matricesGaussian central limit theoremfourth-order kernelsphericity testinghigh-dimensional statisticsquadratic formsrandom matrix theory

0 comments

The pith

Decomposing covariances of quadratic forms into a universal Gaussian part and a fourth-order correction yields explicit Gaussian limits for linear spectral statistics.

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

The paper develops conditions under which linear spectral statistics of high-dimensional sample covariance matrices follow Gaussian distributions even when fourth moments introduce non-universal effects. It starts from a decomposition of the covariance between centered quadratic forms that isolates a universal Gaussian term from a model-dependent fourth-order correction. This decomposition supports an abstract framework that delivers a central limit theorem whose mean and covariance receive explicit adjustments governed by a bilinear fourth-order kernel. The framework is realized through a blockwise mixed radial model in which the correction further splits into entrywise fourth-moment and blockwise energy-fluctuation components, with the latter sometimes changing the fluctuation scale. As a statistical application the same correction reduces to a single scalar under the spherical null, supplying a data-driven adjustment to John's sphericity test.

Core claim

Under the GHOST framework, which rests on a decomposition of the covariance between centered quadratic forms into a universal Gaussian component and a structured fourth-order correction, linear spectral statistics admit a Gaussian central limit theorem whose mean and covariance receive explicit adjustments from a bilinear fourth-order kernel. Boundary examples demonstrate that the stated conditions are nearly necessary for this form of universal Gaussian closure. In the blockwise mixed radial model that verifies the assumptions, the fourth-order correction further decomposes into an entrywise fourth-moment term and a blockwise energy-fluctuation term; the latter can induce a phase transition

What carries the argument

The GHOST framework, an abstract structure for universal Gaussian central limit theorems built on the decomposition of covariances of quadratic forms into a universal Gaussian part and a model-dependent fourth-order correction, together with a bilinear fourth-order kernel that supplies the explicit mean and covariance adjustments.

If this is right

The mean and covariance of linear spectral statistics receive explicit corrections determined by the fourth-order kernel.
Boundary examples indicate the conditions are close to necessary for universal Gaussian behavior.
In the blockwise model the correction splits into entrywise and blockwise components, with the latter possibly changing fluctuation scale.
Under the spherical null the correction reduces to one scalar parameter for a data-driven adjustment to John's test.

Where Pith is reading between the lines

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

The same decomposition approach could apply to other ensembles of quadratic forms arising in high-dimensional statistics beyond sample covariance matrices.
The phase transition in fluctuation scale may change the power of tests that rely on these statistics.
Estimating the single scalar correction parameter from data invites finite-sample analysis for moderate dimensions.

Load-bearing premise

The covariance of centered quadratic forms must decompose into a universal Gaussian part plus a model-dependent fourth-order correction that satisfies the abstract GHOST framework assumptions.

What would settle it

A counterexample model satisfying the blockwise mixed radial structure but producing a non-Gaussian limit for a linear spectral statistic when the fourth-order kernel is nonzero would falsify the central limit theorem.

read the original abstract

We study linear spectral statistics of high dimensional sample covariance matrices in a regime where the empirical spectral distribution remains governed by the classical sample covariance law but the fluctuation theory is nonclassical. Our starting point is a decomposition of the covariance of centered quadratic forms into a universal Gaussian part and a model dependent fourth order correction. This leads to an abstract framework, termed GHOST, for universal Gaussian central limit theorems under structured fourth order effects. Under this framework, we prove a Gaussian central limit theorem for linear spectral statistics, with explicit mean and covariance corrections determined by a bilinear fourth order kernel. Boundary examples show that the conditions are close to necessary for a broad universal Gaussian closure. We then develop a blockwise mixed radial model that verifies the abstract assumptions and makes the correction explicit. The correction splits into an entrywise fourth moment component and a lockwise energy fluctuation component. The latter may change the fluctuation scale, leading to a phase transition at the level of fluctuations. As an application, we study sphericity testing. Under the spherical null, the general correction collapses to a single scalar parameter, yielding a feasible data driven correction of John's test.

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.

Desk Editor's Note private letter to a colleague

read the letter

This paper gives a GHOST framework and explicit radial model for Gaussian CLTs on linear spectral statistics under controlled fourth-order effects, with a clean reduction to a scalar correction for John's sphericity test. They start from a decomposition of the covariance of centered quadratic forms into a universal Gaussian piece plus a model-dependent fourth-order term, then build an abstract setup that yields explicit mean and covariance corrections through a bilinear kernel. The blockwise mixed radial model verifies the assumptions, splits the correction into entrywise fourth-moment and blockwise energy-fluctuation parts, and flags a possible phase transition in fluctuation scale. Boundary examples are offered to suggest the conditions sit close to necessary for this style of universal Gaussian closure. The sphericity application follows directly because the general correction collapses to one scalar parameter under the null, producing a feasible data-driven adjustment to John's test. What is new is the GHOST framework itself, the bilinear kernel construction, the radial model that makes the terms concrete, and the phase-transition observation. The decomposition is derived from the model assumptions rather than reverse-engineered, and the application is direct. The paper is clear on its program and avoids obvious circularity. The main soft spots are that the abstract only outlines the proof strategy and conditions, so the uniformity of the error bounds and the practical reach of the GHOST assumptions remain to be checked in the full derivations. The radial model works by design, but its generality beyond the constructed examples needs scrutiny, and the sharpness of the boundary cases would benefit from more detail. Readers working on high-dimensional covariance inference or non-classical random-matrix fluctuations will get usable tools from this, especially those who want explicit corrections instead of pure existence statements. It deserves a serious referee because the overall argument is coherent and the statistical payoff is concrete. I would recommend sending it to peer review.

Referee Report

1 major / 3 minor

Summary. The manuscript develops the GHOST framework for establishing Gaussian central limit theorems for linear spectral statistics of high-dimensional sample covariance matrices. Starting from a decomposition of the covariance of centered quadratic forms into a universal Gaussian component and a model-dependent fourth-order correction, the authors prove a CLT with explicit mean and covariance corrections determined by a bilinear fourth-order kernel. They introduce a blockwise mixed radial model to realize the framework explicitly, where the correction splits into entrywise fourth-moment and blockwise energy-fluctuation terms, potentially inducing a fluctuation scale phase transition. Boundary examples are provided to show the conditions are near-optimal, and an application to sphericity testing yields a corrected version of John's test under the spherical null.

Significance. If the technical details hold, this work advances the understanding of fluctuation theory beyond classical regimes in random matrix theory by incorporating structured fourth-order effects in a controlled manner. The explicit nature of the corrections and the construction of a verifying model are significant strengths, as is the reduction of the sphericity test to a scalar parameter correction. This could lead to improved statistical procedures in high-dimensional data analysis. The near-optimality via boundary examples adds rigor to the universality claim.

major comments (1)

§4 (blockwise mixed radial model): the claim that the covariance decomposition yields an explicit bilinear kernel correction requires explicit remainder bounds showing that the fourth-order term does not interfere with the Gaussian limit; without these, the phase-transition statement in the fluctuation scale remains conditional on unverified decay rates.

minor comments (3)

Abstract: 'lockwise' is a typographical error for 'blockwise'.
Notation: the bilinear fourth-order kernel is introduced without a dedicated display equation in the framework section; adding one would improve readability when referring to the mean/covariance corrections.
The sphericity-testing application reduces John's statistic to a scalar correction, but the feasible estimator for that scalar is only sketched; a short appendix with the explicit plug-in formula would strengthen the statistical claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thorough review and constructive feedback on our manuscript. We appreciate the positive assessment of the GHOST framework and its applications. We address the major comment below.

read point-by-point responses

Referee: §4 (blockwise mixed radial model): the claim that the covariance decomposition yields an explicit bilinear kernel correction requires explicit remainder bounds showing that the fourth-order term does not interfere with the Gaussian limit; without these, the phase-transition statement in the fluctuation scale remains conditional on unverified decay rates.

Authors: We acknowledge the referee's observation that explicit remainder bounds would strengthen the argument in Section 4. The blockwise mixed radial model is constructed so that the GHOST assumptions hold with decay rates on fourth-order moments and blockwise energy fluctuations that force the remainder in the covariance decomposition to be o(1). Nevertheless, to remove any ambiguity, we will insert a dedicated lemma in the revision that derives explicit bounds on these remainder terms directly from the radial structure and blockwise independence. These bounds will confirm that the fourth-order correction remains negligible relative to the Gaussian term and will make the conditions for the fluctuation-scale phase transition fully explicit rather than implicit in the model assumptions. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The derivation begins with an explicit starting decomposition of the covariance of centered quadratic forms into a universal Gaussian part plus a model-dependent fourth-order correction; this is taken as given rather than derived from the target CLT. The abstract GHOST framework is then defined from these assumptions, and the Gaussian CLT for linear spectral statistics is proved under the framework, yielding explicit mean and covariance corrections via a bilinear fourth-order kernel. A blockwise mixed radial model is subsequently constructed to satisfy the abstract assumptions and render the correction explicit (splitting into entrywise fourth-moment and blockwise energy-fluctuation terms). Boundary examples are used only to argue near-necessity of the conditions, not to presuppose the result. The sphericity-testing application follows by direct specialization of the general correction to a scalar under the spherical null. No equation or step reduces the claimed CLT or corrections to the inputs by construction, and no load-bearing self-citation or uniqueness theorem is invoked.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

The central claims rest on a new abstract framework and a new model whose independent verification is not supplied in the abstract; the fourth-order kernel functions as a model-specific parameter that determines the corrections.

free parameters (1)

bilinear fourth order kernel
Determines the explicit mean and covariance corrections; its form is model-dependent and not derived from first principles within the abstract.

axioms (1)

domain assumption Covariance of centered quadratic forms decomposes into a universal Gaussian part plus a model-dependent fourth-order correction
This decomposition is the explicit starting point for the GHOST framework.

invented entities (2)

GHOST framework no independent evidence
purpose: Abstract conditions guaranteeing universal Gaussian CLTs under structured fourth-order effects
Newly introduced abstract framework in the paper.
blockwise mixed radial model no independent evidence
purpose: Concrete model that satisfies GHOST assumptions and renders the fourth-order correction explicit, including the phase transition
New model constructed to verify the abstract framework.

pith-pipeline@v0.9.0 · 5500 in / 1530 out tokens · 61035 ms · 2026-05-13T02:05:49.914153+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
Our starting point is a decomposition of the covariance of centered quadratic forms into a universal Gaussian part and a model dependent fourth order correction. This leads to an abstract framework, termed GHOST... bilinear fourth order kernel.
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
Theorem 3.2... asymptotically Gaussian with... Mn(f) = M0,n(f) + M1,n(f) ... Vn(f,g) = V0,n + V1,n driven by 1/n Γn(H^{-1},H^{-1})

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

[1]

barticle [author] Bai , Z. D. Z. D. Silverstein , Jack W. J. W. ( 1998 ). No eigenvalues outside the support of the limiting spectral distribution of large-dimensional sample covariance matrices . The Annals of Probability 26 316--345 . 10.1214/aop/1022855421 barticle

work page doi:10.1214/aop/1022855421 1998
[2]

barticle [author] Bai , Z. D. Z. D. Silverstein , Jack W. J. W. ( 2004 ). Clt for linear spectral statistics of large-dimensional sample covariance matrices . Annals of Probability 32 553--605 . 10.1214/aop/1078415845 barticle

work page doi:10.1214/aop/1078415845 2004
[3]

Springer, New York (2010).https://doi.org/10.1007/978-1-4419-0661-8

bbook [author] Bai , Zhidong Z. Silverstein , Jack W. J. W. ( 2010 ). Spectral Analysis of Large Dimensional Random Matrices . Springer Series in Statistics . Springer New York , New York, NY . 10.1007/978-1-4419-0661-8 bbook

work page doi:10.1007/978-1-4419-0661-8 2010
[4]

, Pan , Guangming G

barticle [author] Bao , Zhigang Z. , Pan , Guangming G. Zhou , Wang W. ( 2015 ). Asymptotic mutual information statistics of MIMO channels and CLT of sample covariance matrices . IEEE Transactions on Information Theory 61 3413--3426 . 10.1109/TIT.2015.2421894 barticle

work page doi:10.1109/tit.2015.2421894 2015
[5]

, Lin , Liang Ching L

barticle [author] Bao , Zhigang Z. , Lin , Liang Ching L. C. , Pan , Guangming G. Zhou , Wang W. ( 2015 ). Spectral statistics of large dimensional spearman's rank correlation matrix and its application . Annals of Statistics 43 2588--2623 . 10.1214/15-AOS1353 barticle

work page doi:10.1214/15-aos1353 2015
[6]

, Hu , Jiang J

barticle [author] Bao , Zhigang Z. , Hu , Jiang J. , Xu , Xiaocong X. Zhang , Xiaozhuo X. ( 2024 ). Spectral statistics of sample block correlation matrices . The Annals of Statistics 52 1873--1898 . barticle

work page 2024
[7]

, Han , Xiao X

barticle [author] Gao , Jiti J. , Han , Xiao X. , Pan , Guangming G. Yang , Yanrong Y. ( 2017 ). High dimensional correlation matrices: the central limit theorem and its applications . Journal of the Royal Statistical Society. Series B: Statistical Methodology 79 677--693 . 10.1111/rssb.12189 barticle

work page doi:10.1111/rssb.12189 2017
[8]

, Yang , Qing Q

barticle [author] Hu , Yanlin Y. , Yang , Qing Q. Han , Xiao X. ( 2026 ). Generalized linear spectral statistics of high-dimensional sample covariance matrices and its applications . The Annals of Statistics 54 1028--1052 . barticle

work page 2026
[9]

, Li , Weiming W

barticle [author] Hu , Jiang J. , Li , Weiming W. , Liu , Zhi Z. Zhou , Wang W. ( 2019 ). High-dimensional covariance matrices in elliptical distributions with application to spherical test . Annals of Statistics 47 527--555 . 10.1214/18-AOS1699 barticle

work page doi:10.1214/18-aos1699 2019
[10]

Yao , Jianfeng J

barticle [author] Li , Zeng Z. Yao , Jianfeng J. ( 2016 ). Testing the sphericity of a covariance matrix when the dimension is much larger than the sample size . Electronic Journal of Statistics 10 2973--3010 . 10.1214/16-EJS1199 barticle

work page doi:10.1214/16-ejs1199 2016
[11]

Yao , Jianfeng J

barticle [author] Li , Weiming W. Yao , Jianfeng J. ( 2018 ). On structure testing for component covariance matrices of a high dimensional mixture . Journal of the Royal Statistical Society. Series B: Statistical Methodology 80 293--318 . 10.1111/rssb.12248 barticle

work page doi:10.1111/rssb.12248 2018
[12]

, Wang , Qinwen Q

barticle [author] Li , Weiming W. , Wang , Qinwen Q. , Yao , Jianfeng J. Zhou , Wang W. ( 2021 ). On eigenvalues of a high-dimensional spatial-sign covariance matrix . Bernoulli 28 606--637 . 10.3150/21-BEJ1360 barticle

work page doi:10.3150/21-bej1360 2021
[13]

, Pan , Guangming G

barticle [author] Li , Huiqin H. , Pan , Guangming G. , Yin , Yanqing Y. Zhou , Wang W. ( 2024 a). Spectral analysis of gram matrices with missing at random observations: Convergence, central limit theorems, and applications in statistical inference . The Annals of Statistics 52 1254--1275 . barticle

work page 2024
[14]

, Pan , Guangming G

barticle [author] Li , Yanpeng Y. , Pan , Guangming G. , Xie , Jiahui J. Zhou , Wang W. ( 2024 b). Necessary and sufficient condition for CLT of linear spectral statistics of sample correlation matrices . arXiv preprint arXiv:2409.12536 . barticle

work page arXiv 2024
[15]

, Hu , Jiang J

barticle [author] Liu , Zhijun Z. , Hu , Jiang J. , Bai , Zhidong Z. Song , Haiyan H. ( 2023 ). A CLT for the LSS of large-dimensional sample covariance matrices with diverging spikes . The Annals of Statistics 51 2246--2271 . barticle

work page 2023
[16]

, Moreira , Marcelo J

barticle [author] Onatski , Alexei A. , Moreira , Marcelo J. M. J. Hallin , Marc M. ( 2013 ). Asymptotic power of sphericity tests for high-dimensional data . Annals of Statistics 41 1204--1231 . 10.1214/13-AOS1100 barticle

work page doi:10.1214/13-aos1100 2013
[17]

, Li , Zeng Z

barticle [author] Qiu , Jiaxin J. , Li , Zeng Z. Yao , Jianfeng J. ( 2023 ). Asymptotic normality for eigenvalue statistics of a general sample covariance matrix when p/n and applications . Annals of Statistics 51 1427--1451 . 10.1214/23-AOS2300 barticle

work page doi:10.1214/23-aos2300 2023
[18]

( 2018 )

bbook [author] Vershynin , Roman R. ( 2018 ). High-dimensional probability: An introduction with applications in data science 47 . Cambridge university press . bbook

work page 2018
[19]

Yao , Jianfeng J

barticle [author] Wang , Qinwen Q. Yao , Jianfeng J. ( 2013 ). On the sphericity test with large-dimensional observations . Electronic Journal of Statistics 7 2164--2192 . 10.1214/13-EJS842 barticle

work page doi:10.1214/13-ejs842 2013
[20]

barticle [author] Yang , By Yanrong B. Y. Pan , Guangming G. ( 2015 ). Independence test for high dimensional data based on regularized canonical correlation coefficients . Annals of Statistics 43 467--500 . 10.1214/14-AOS1284 barticle

work page doi:10.1214/14-aos1284 2015
[21]

, Bai , Zhi-Dong Z.-D

barticle [author] Yin , Yong-Qua Y.-Q. , Bai , Zhi-Dong Z.-D. Krishnaiah , Pathak R P. R. ( 1988 ). On the limit of the largest eigenvalue of the large dimensional sample covariance matrix . Probability theory and related fields 78 509--521 . barticle

work page 1988
[22]

, Zheng , Shurong S

barticle [author] Yin , Yanqing Y. , Zheng , Shurong S. Zou , Tingting T. ( 2023 ). Central Limit Theorem of Linear Spectral Statistics of High-dimensional Sample Correlation Matrices . Bernoulli 2 . barticle

work page 2023
[23]

Zhou , Wang W

barticle [author] Yin , Yanqing Y. Zhou , Wang W. ( 2026 ). Supplementary material for ``The geometry of spectral fluctuations: On near-optimal conditions for universal Gaussian CLTs, with statistical applications'' . barticle

work page 2026
[24]

, Zheng , Shurong S

barticle [author] Zhang , Zhixiang Z. , Zheng , Shurong S. , Pan , Guangming G. Zhong , Ping-Shou P.-S. ( 2022 ). Asymptotic independence of spiked eigenvalues and linear spectral statistics for large sample covariance matrices . The Annals of Statistics 50 2205--2230 . 10.1214/22-AOS2183 barticle

work page doi:10.1214/22-aos2183 2022
[25]

, Bai , Zhidong Z

barticle [author] Zheng , Shurong S. , Bai , Zhidong Z. Yao , Jianfeng J. ( 2015 ). Substitution principle for CLT of linear spectral statistics of high-dimensional sample covariance matrices with applications to hypothesis testing . Annals of Statistics 43 546--591 . 10.1214/14-AOS1292 barticle

work page doi:10.1214/14-aos1292 2015
[26]

, Peng , Liuhua L

barticle [author] Zou , Changliang C. , Peng , Liuhua L. , Feng , Long L. Wang , Zhaojun Z. ( 2014 ). Multivariate sign-based high-dimensional tests for sphericity . Biometrika 101 229--236 . 10.1093/biomet/ast040 barticle

work page doi:10.1093/biomet/ast040 2014
[27]

Zhou , Wang W

barticle [author] Bai , Zhidong Z. Zhou , Wang W. ( 2008 ). Large sample covariance matrices without independence structures in columns . Statistica Sinica 18 425--442 . barticle

work page 2008
[28]

bbook [author] Kingman , J. F. C. J. F. C. Billingsley , P. P. ( 1969 ). Convergence of Probability Measures. 18 . Wiley-Interscience . 10.2307/2346589 bbook

work page doi:10.2307/2346589 1969
[29]

barticle [author] Silverstein , Jack W. J. W. ( 1995 ). Strong convergence of the empirical distribution of eigenvalues of large dimensional random matrices . Journal of Multivariate Analysis 55 331--339 . 10.1006/jmva.1995.1083 barticle

work page doi:10.1006/jmva.1995.1083 1995
[30]

Ma , Yanyuan Y

barticle [author] Yin , Yanqing Y. Ma , Yanyuan Y. ( 2022 ). Properties of Eigenvalues and Eigenvectors of Large-Dimensional Sample Correlation Matrices . Annals of Applied Probability 32 4763--4802 . 10.1214/22-AAP1802 barticle

work page doi:10.1214/22-aap1802 2022