arxiv: 2604.18181 · v2 · submitted 2026-04-20 · 🧮 math.ST · stat.TH

Recognition: unknown

Spectral approximation for the separable covariance mixture model

Ben Deitmar

Authors on Pith no claims yet

Pith reviewed 2026-05-10 03:37 UTC · model grok-4.3

classification 🧮 math.ST stat.TH

keywords separable covariance mixture modelresolvent approximationdeterministic equivalentsample covariance matrixlimiting spectral distributionnon-asymptotic boundsrandom matrix theory

0 comments

The pith

Resolvents of the sample covariance matrices approximate deterministic matrices in the separable covariance mixture model.

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

The paper introduces the separable covariance mixture model, in which a data matrix takes the form of a sum of products of fixed matrices with a single random matrix having independent centered unit-variance entries. It establishes that the resolvents of the associated sample covariance matrices approximate explicit deterministic expressions involving weighted sums of the fixed matrices, with weights given by solutions to a dual system of equations. This approximation is non-asymptotic and holds without the simultaneous diagonalizability assumption required in prior work. Such results matter for deriving the limiting spectral distributions of these covariance matrices in high-dimensional data analysis.

Core claim

For Y formed as the sum from r=1 to R of A_r X B_r, the resolvent (1/n Y Y^* - z Id_d)^{-1} approximates -1/z (Id_d + sum_{r,s} delta^{(B)}_{r,s}(z) A_r A_s^*)^{-1}, and similarly for the other resolvent, where delta^{(A)} and delta^{(B)} are the unique solutions to a certain dual system of equations. The results do not require simultaneous diagonalizability of the families of A matrices or B matrices.

What carries the argument

The dual system of equations solved by the R by R matrices delta^{(A)}(z) and delta^{(B)}(z), which parametrize the deterministic equivalents of the resolvents.

If this is right

The approximation enables spectral analysis of the sample covariance without assuming the A_r commute or are simultaneously diagonalizable.
An asymptotic consequence is the limiting spectral distribution of 1/n Y Y^* or 1/n Y^* Y.
The non-asymptotic nature allows finite-dimensional bounds on the approximation error.

Where Pith is reading between the lines

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

This could facilitate parameter estimation in covariance mixture models by matching observed spectra to the deterministic forms.
The approach may generalize to other structured random matrix ensembles in high-dimensional statistics.
Numerical solution of the dual equations offers a practical way to compute approximate eigenvalues for large data matrices.

Load-bearing premise

The existence and uniqueness of solutions to the dual system of equations for delta^{(A)} and delta^{(B)}, assuming the random matrix X has independent centered entries with unit variance.

What would settle it

A direct numerical comparison for moderate dimensions where the norm of the difference between the actual resolvent and the proposed deterministic matrix remains large despite the model assumptions being satisfied.

Figures

Figures reproduced from arXiv: 2604.18181 by Ben Deitmar.

**Figure 1.** Figure 1: Histograms (blue) of the eigenvalues of S˜ for Example 1 with n = 100 (left), Example 2 with n = 200 and R = 4 (middle) and Example 3 with n = 100 and R = 4 (right). Plots (orange) of the maps x 7→ 1 π Im(sν(x + iη)) for η = 0.05 are overlaid, where ν is as defined in Theorem 1.1 [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗

**Figure 2.** Figure 2: Average approximation errors from (2.8), where the upper absolute value in (2.8) is called "A-Error" and the lower absolute value in (2.8) is called "B-Error". The test-matrices M, M˜ are as defined in Remark 2.7. The data is from 25 realizations of Examples 1 (left), 2 (middle) and 3 (right) with n ranging from 10 to 100. The z-points are z1 = 1.5 + i and z2 = 1.5 + 0.1i (left), z1 = 6 + 0.1i and z2 = 0.5… view at source ↗

read the original abstract

This paper introduces the separable covariance mixture model, which assumes a data-matrix $Y$ to be of the form $$ \sum\limits_{r=1}^R A_r X B_r $$ for one random $(d \times n)$-matrix $X$ with independent centered variance-one entries, and for two families of deterministic matrices $A_1,\dots,A_R \in \mathbb{C}^{d \times d}$ and $B_1,\dots,B_R \in \mathbb{C}^{n \times n}$. Under certain assumptions, it is shown that the resolvents $(\frac{1}{n} Y Y^* - z \operatorname{Id}_d)^{-1}$ and $(\frac{1}{n} Y^* Y - z \operatorname{Id}_n)^{-1}$ respectively approximate the deterministic matrices $$ -\frac{1}{z}\Big( \operatorname{Id}_d + \sum\limits_{r,s=1}^R \delta^{(B)}_{r,s}(z) A_{r} A_{s}^* \Big)^{-1} \ \ \text{ and } \ \ -\frac{1}{z}\Big( \operatorname{Id}_n + \sum\limits_{r,s=1}^R \delta^{(A)}_{r,s}(z) B_{s}^*B_{r} \Big)^{-1} \ , $$ where $\delta^{(A)}, \delta^{(B)} \in \mathbb{C}^{R \times R}$ are uniquely defined solutions to a certain dual system of equations. The results are non-asymptotic and do not require simultaneous diagonalizability of the families $(A_r)_{r \leq R}$ or $(B_r)_{r \leq R}$, as was required in previous works such as [Hazarika and Paul (2025)] or [Mei et al. (2023)]. An asymptotic application, which describes the limiting spectral distribution of the sample covariance matrix analogues $\frac{1}{n} Y Y^*$ or $\frac{1}{n} Y^* Y$, is included.

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

The paper gives a non-asymptotic resolvent approximation for the separable covariance mixture that drops the simultaneous diagonalizability requirement, but the uniqueness of the delta solutions is stated rather than derived in detail.

read the letter

The main takeaway is that this work extends spectral approximations to the model Y = sum A_r X B_r where X has iid centered unit-variance entries. It supplies deterministic equivalents for the resolvents of the sample covariances without forcing the A family or the B family to be simultaneously diagonalizable, which earlier papers needed. The deltas are defined as the unique solution to a dual system, and the result is non-asymptotic with an application to limiting spectral distributions of the sample covariances.

Referee Report

3 major / 2 minor

Summary. The paper introduces the separable covariance mixture model, in which a data matrix Y takes the form sum_{r=1}^R A_r X B_r with X having i.i.d. centered unit-variance entries and A_r, B_r deterministic. It claims that, under certain assumptions, the resolvents of (1/n) Y Y^* and (1/n) Y^* Y admit non-asymptotic deterministic approximations expressed via matrices delta^{(A)}(z) and delta^{(B)}(z) that are the unique solutions to a dual system of equations; the approximations do not require simultaneous diagonalizability of the families (A_r) or (B_r). An asymptotic corollary on the limiting spectral distribution of the sample covariance matrices is also derived.

Significance. If the existence and uniqueness of the delta solutions can be established under explicit, verifiable conditions on the A_r, B_r and z, the result would extend random-matrix techniques to a broader class of structured covariance models without the diagonalizability restrictions of prior work. The non-asymptotic character and the dual-system formulation are potentially useful for applications in high-dimensional statistics and signal processing.

major comments (3)

[Main theorem / dual system definition] Main theorem (likely §3 or §4): the statement that delta^{(A)}, delta^{(B)} are 'uniquely defined solutions' to the dual system is invoked to make the deterministic equivalents well-defined, yet the manuscript supplies no explicit conditions on the A_r, B_r or on Im z > 0 that guarantee existence and uniqueness. The model assumption on X is used only for the stochastic approximation step; without a separate lemma establishing well-posedness of the fixed-point system, the central claim is not fully supported.
[Resolvent approximation theorem] Theorem on resolvent approximation: the error bounds between the random resolvents and the displayed deterministic matrices are stated to be non-asymptotic, but the precise dependence of the error on n, d, R, and the operator norms of the A_r, B_r is not made explicit in the abstract or theorem statement. This makes it difficult to assess the practical range of validity.
[Asymptotic corollary] Asymptotic application (limiting spectral distribution): the passage from the non-asymptotic resolvent approximation to the LSD of (1/n)YY^* relies on the deltas converging to deterministic limits; the manuscript does not verify that the dual system admits a unique solution in the large-n,d regime or that the convergence of deltas is uniform in z.

minor comments (2)

[Introduction / Main result] Notation for the dual system: the precise form of the equations satisfied by delta^{(A)} and delta^{(B)} should be displayed explicitly (rather than described as 'a certain dual system') already in the introduction or statement of the main result.
[Introduction] Comparison with prior work: the claim that simultaneous diagonalizability is no longer required is important, but a short paragraph contrasting the new assumptions with those of Hazarika-Paul (2025) and Mei et al. (2023) would help readers.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thorough review and insightful comments on our manuscript. We appreciate the recognition of the potential extensions to random matrix techniques for structured covariance models. We address each of the major comments below, indicating the revisions we will make to strengthen the paper.

read point-by-point responses

Referee: Main theorem (likely §3 or §4): the statement that delta^{(A)}, delta^{(B)} are 'uniquely defined solutions' to the dual system is invoked to make the deterministic equivalents well-defined, yet the manuscript supplies no explicit conditions on the A_r, B_r or on Im z > 0 that guarantee existence and uniqueness. The model assumption on X is used only for the stochastic approximation step; without a separate lemma establishing well-posedness of the fixed-point system, the central claim is not fully supported.

Authors: We acknowledge this gap in the presentation. Although the existence and uniqueness follow from the assumptions on the boundedness of the operator norms of the A_r and B_r and the positive imaginary part of z (as implicitly used in the contraction arguments in the proofs), we agree that an explicit statement is necessary. In the revised manuscript, we will add a dedicated lemma (new Lemma 3.1) proving the existence and uniqueness of the solutions to the dual system of equations under these conditions, using a fixed-point theorem. This will support the central claim without relying solely on the stochastic approximation step. revision: yes
Referee: Theorem on resolvent approximation: the error bounds between the random resolvents and the displayed deterministic matrices are stated to be non-asymptotic, but the precise dependence of the error on n, d, R, and the operator norms of the A_r, B_r is not made explicit in the abstract or theorem statement. This makes it difficult to assess the practical range of validity.

Authors: We thank the referee for this observation. The error bounds in the theorem do depend explicitly on n, d, R, and the maximum operator norms of the A_r and B_r, as derived in the proof via concentration inequalities and matrix perturbation bounds. To improve clarity, we will revise the theorem statement to explicitly include this dependence (with the constants made visible) and add a remark on the practical range of validity. revision: yes
Referee: Asymptotic application (limiting spectral distribution): the passage from the non-asymptotic resolvent approximation to the LSD of (1/n)YY^* relies on the deltas converging to deterministic limits; the manuscript does not verify that the dual system admits a unique solution in the large-n,d regime or that the convergence of deltas is uniform in z.

Authors: We agree that the asymptotic corollary requires additional justification for the convergence. In the revision, we will include a new proposition showing that, under the same assumptions with n,d → ∞, the solutions δ^{(A)}(z), δ^{(B)}(z) converge to the unique solutions of the limiting system, with the convergence being uniform on compact subsets of the upper half-plane. This will be based on the continuity of the fixed-point map and the non-asymptotic error bounds, thereby rigorously justifying the limiting spectral distribution result. revision: yes

Circularity Check

0 steps flagged

No circularity: deltas defined as solutions to derived fixed-point system, approximation proven independently

full rationale

The paper states that under certain assumptions the resolvents approximate the displayed deterministic matrices, with deltas uniquely solving a dual system of equations. This system is presented as part of the result rather than presupposed by definition or fitted to data. No equations reduce the claimed approximation to its own inputs by construction, no parameters are fitted then renamed as predictions, and no load-bearing self-citations or imported uniqueness theorems appear. The non-asymptotic claim rests on the model assumptions for X and the existence/uniqueness of deltas, but these are external to the derivation chain itself and do not create a self-referential loop. The result is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The claim rests on standard properties of matrix resolvents and the existence/uniqueness of solutions to the dual fixed-point system for the deltas; no free parameters are fitted to data and no new entities are postulated.

axioms (2)

domain assumption Existence and uniqueness of solutions to the dual system of equations defining delta^{(A)} and delta^{(B)}
Invoked to guarantee the deterministic approximants are well-defined; appears in the statement of the main result.
domain assumption X has independent centered entries of variance one
Core model assumption used to control the fluctuation of the random matrix and derive the deterministic equivalent.

pith-pipeline@v0.9.0 · 5674 in / 1492 out tokens · 37099 ms · 2026-05-10T03:37:10.511050+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

162 extracted references · 77 canonical work pages

[1]

Fleermann, Michael and Kirsch, Werner , TITLE =. Probab. Surv. , FJOURNAL =. 2023 , PAGES =. doi:10.1214/23-ps16 , URL =

work page doi:10.1214/23-ps16 2023
[2]

Matematicheskii Sbornik , FJOURNAL =

Marchenko, Vladimir Alexandrovich and Pastur, Leonid Andreevich , TITLE =. Matematicheskii Sbornik , FJOURNAL =. 1967 , PAGES =. doi:10.1070/SM1967v001n04ABEH001994 , URL =

work page doi:10.1070/sm1967v001n04abeh001994 1967
[3]

Yin, Y. Q. , TITLE =. J. Multivariate Anal. , FJOURNAL =. 1986 , NUMBER =. doi:10.1016/0047-259X(86)90019-9 , URL =

work page doi:10.1016/0047-259x(86)90019-9 1986
[4]

and Bai, Z

Silverstein, Jack W. and Bai, Z. D. , TITLE =. J. Multivariate Anal. , FJOURNAL =. 1995 , NUMBER =. doi:10.1006/jmva.1995.1051 , URL =

work page doi:10.1006/jmva.1995.1051 1995
[5]

Bai, Zhidong and Zhou, Wang , TITLE =. Statist. Sinica , FJOURNAL =. 2008 , NUMBER =

2008
[6]

Stochastic Process

Fleermann, Michael and Heiny, Johannes , TITLE =. Stochastic Process. Appl. , FJOURNAL =. 2023 , PAGES =. doi:10.1016/j.spa.2023.04.020 , URL =

work page doi:10.1016/j.spa.2023.04.020 2023
[7]

Electron

Yaskov, Pavel , TITLE =. Electron. Commun. Probab. , FJOURNAL =. 2016 , PAGES =. doi:10.1214/16-ECP4748 , URL =

work page doi:10.1214/16-ecp4748 2016
[8]

Dong, Zhaorui and Yao, Jianfeng , TITLE =. Statist. Probab. Lett. , FJOURNAL =. 2025 , PAGES =. doi:10.1016/j.spl.2025.110377 , URL =

work page doi:10.1016/j.spl.2025.110377 2025
[9]

D\"ornemann, Nina and Heiny, Johannes , TITLE =. Ann. Appl. Probab. , FJOURNAL =. 2025 , NUMBER =. doi:10.1214/25-AAP2181 , URL =

work page doi:10.1214/25-aap2181 2025
[10]

Ajanki, Oskari H. and Erd. Stability of the matrix. Probab. Theory Related Fields , FJOURNAL =. 2019 , NUMBER =. doi:10.1007/s00440-018-0835-z , URL =

work page doi:10.1007/s00440-018-0835-z 2019
[11]

Bai, Z. D. and Silverstein, Jack W. , TITLE =. Ann. Probab. , FJOURNAL =. 2004 , NUMBER =. doi:10.1214/aop/1078415845 , URL =

work page doi:10.1214/aop/1078415845 2004
[12]

D\"ornemann, Nina and Dette, Holger , TITLE =. Ann. Inst. Henri Poincar\'e. 2024 , NUMBER =. doi:10.1214/22-aihp1339 , URL =

work page doi:10.1214/22-aihp1339 2024
[13]

Qiu, Jiaxin and Li, Zeng and Yao, Jianfeng , TITLE =. Ann. Statist. , FJOURNAL =. 2023 , NUMBER =. doi:10.1214/23-aos2300 , URL =

work page doi:10.1214/23-aos2300 2023
[14]

Li, Weiming and Li, Zeng and Yao, Jianfeng , TITLE =. Scand. J. Stat. , FJOURNAL =. 2018 , NUMBER =. doi:10.1111/sjos.12320 , URL =

work page doi:10.1111/sjos.12320 2018
[15]

Yin, Y. Q. and Bai, Z. D. and Krishnaiah, P. R. , TITLE =. Probab. Theory Related Fields , FJOURNAL =. 1988 , NUMBER =. doi:10.1007/BF00353874 , URL =

work page doi:10.1007/bf00353874 1988
[16]

Najim, Jamal and Yao, Jianfeng , TITLE =. Ann. Appl. Probab. , FJOURNAL =. 2016 , NUMBER =. doi:10.1214/15-AAP1135 , URL =

work page doi:10.1214/15-aap1135 2016
[17]

Isotropic local laws for sample covariance and generalized

Bloemendal, Alex and Erd. Isotropic local laws for sample covariance and generalized. Electron. J. Probab. , FJOURNAL =. 2014 , PAGES =. doi:10.1214/ejp.v19-3054 , URL =

work page doi:10.1214/ejp.v19-3054 2014
[18]

Knowles, Antti and Yin, Jun , TITLE =. Probab. Theory Related Fields , FJOURNAL =. 2017 , NUMBER =. doi:10.1007/s00440-016-0730-4 , URL =

work page doi:10.1007/s00440-016-0730-4 2017
[19]

Bloemendal, Alex and Knowles, Antti and Yau, Horng-Tzer and Yin, Jun , TITLE =. Probab. Theory Related Fields , FJOURNAL =. 2016 , NUMBER =. doi:10.1007/s00440-015-0616-x , URL =

work page doi:10.1007/s00440-015-0616-x 2016
[20]

Hwang, Jong Yun and Lee, Ji Oon and Schnelli, Kevin , TITLE =. Ann. Appl. Probab. , FJOURNAL =. 2019 , NUMBER =. doi:10.1214/19-AAP1472 , URL =

work page doi:10.1214/19-aap1472 2019
[21]

Jonsson, Dag , TITLE =. J. Multivariate Anal. , FJOURNAL =. 1982 , NUMBER =. doi:10.1016/0047-259X(82)90080-X , URL =

work page doi:10.1016/0047-259x(82)90080-x 1982
[22]

El Karoui, Noureddine , TITLE =. Ann. Statist. , FJOURNAL =. 2008 , NUMBER =. doi:10.1214/07-AOS581 , URL =

work page doi:10.1214/07-aos581 2008
[23]

Li, Weiming and Chen, Jiaqi and Qin, Yingli and Bai, Zhidong and Yao, Jianfeng , TITLE =. J. Statist. Plann. Inference , FJOURNAL =. 2013 , NUMBER =. doi:10.1016/j.jspi.2013.06.017 , URL =

work page doi:10.1016/j.jspi.2013.06.017 2013
[24]

2015 , PAGES =

Yao, Jianfeng and Zheng, Shurong and Bai, Zhidong , TITLE =. 2015 , PAGES =. doi:10.1017/CBO9781107588080 , URL =

work page doi:10.1017/cbo9781107588080 2015
[25]

Bai, Zhidong and Chen, Jiaqi and Yao, Jianfeng , TITLE =. Aust. N. Z. J. Stat. , FJOURNAL =. 2010 , NUMBER =. doi:10.1111/j.1467-842X.2010.00590.x , URL =

work page doi:10.1111/j.1467-842x.2010.00590.x 2010
[26]

Arizmendi, Octavio and Tarrago, Pierre and Vargas, Carlos , TITLE =. Ann. Inst. Henri Poincar\'e. 2020 , NUMBER =. doi:10.1214/20-AIHP1050 , URL =

work page doi:10.1214/20-aihp1050 2020
[27]

Kong, Weihao and Valiant, Gregory , TITLE =. Ann. Statist. , FJOURNAL =. 2017 , NUMBER =. doi:10.1214/16-AOS1525 , URL =

work page doi:10.1214/16-aos1525 2017
[28]

Donoho, David and Gavish, Matan and Johnstone, Iain , TITLE =. Ann. Statist. , FJOURNAL =. 2018 , NUMBER =. doi:10.1214/17-AOS1601 , URL =

work page doi:10.1214/17-aos1601 2018
[29]

Ledoit, Olivier and Wolf, Michael , TITLE =. Ann. Statist. , FJOURNAL =. 2012 , NUMBER =. doi:10.1214/12-AOS989 , URL =

work page doi:10.1214/12-aos989 2012
[30]

Ledoit, Olivier and Wolf, Michael , TITLE =. J. Multivariate Anal. , FJOURNAL =. 2015 , PAGES =. doi:10.1016/j.jmva.2015.04.006 , URL =

work page doi:10.1016/j.jmva.2015.04.006 2015
[31]

Mei, Tianxing and Wang, Chen and Yao, Jianfeng , TITLE =. Ann. Statist. , FJOURNAL =. 2023 , NUMBER =. doi:10.1214/23-aos2263 , URL =

work page doi:10.1214/23-aos2263 2023
[32]

Liu, Haoyang and Aue, Alexander and Paul, Debashis , TITLE =. Ann. Statist. , FJOURNAL =. 2015 , NUMBER =. doi:10.1214/14-AOS1294 , URL =

work page doi:10.1214/14-aos1294 2015
[33]

Jin, Baisuo and Wang, Cheng and Miao, Baiqi and Lo Huang, Mong-Na , TITLE =. J. Multivariate Anal. , FJOURNAL =. 2009 , NUMBER =. doi:10.1016/j.jmva.2009.06.011 , URL =

work page doi:10.1016/j.jmva.2009.06.011 2009
[34]

Yao, Jianfeng , TITLE =. Statist. Probab. Lett. , FJOURNAL =. 2012 , NUMBER =. doi:10.1016/j.spl.2011.08.011 , URL =

work page doi:10.1016/j.spl.2011.08.011 2012
[35]

Bhattacharjee, Monika and Bose, Arup , TITLE =. Ann. Statist. , FJOURNAL =. 2016 , NUMBER =. doi:10.1214/15-AOS1378 , URL =

work page doi:10.1214/15-aos1378 2016
[36]

Ding, Yi and Zheng, Xinghua , TITLE =. Ann. Statist. , FJOURNAL =. 2024 , NUMBER =. doi:10.1214/24-aos2381 , URL =

work page doi:10.1214/24-aos2381 2024
[37]

39th Annual Meeting IMS, Atlanta, GA, 1975 , year=

Estimation of a covariance matrix , author=. 39th Annual Meeting IMS, Atlanta, GA, 1975 , year=

1975
[38]

Journal of Soviet Mathematics , volume=

Lectures on the theory of estimation of many parameters , author=. Journal of Soviet Mathematics , volume=. 1986 , publisher=

1986
[39]

Ledoit, Olivier and Wolf, Michael , TITLE =. Ann. Statist. , FJOURNAL =. 2020 , NUMBER =. doi:10.1214/19-AOS1921 , URL =

work page doi:10.1214/19-aos1921 2020
[40]

Journal of Financial Econometrics , volume=

The power of (non-) linear shrinking: A review and guide to covariance matrix estimation , author=. Journal of Financial Econometrics , volume=. 2022 , publisher=

2022
[41]

2024 , note=

Eigenvector distributions and optimal shrinkage estimators for large covariance and precision matrices , author=. 2024 , note=

2024
[42]

, keywords =

Silverstein, Jack W. and Choi, Sang-Il , TITLE =. J. Multivariate Anal. , FJOURNAL =. 1995 , NUMBER =. doi:10.1006/jmva.1995.1058 , URL =

work page doi:10.1006/jmva.1995.1058 1995
[43]

Random Matrices Theory Appl

Dobriban, Edgar , TITLE =. Random Matrices Theory Appl. , FJOURNAL =. 2015 , NUMBER =. doi:10.1142/S2010326315500197 , URL =

work page doi:10.1142/s2010326315500197 2015
[44]

Xia, Ningning and Zheng, Xinghua , TITLE =. Ann. Statist. , FJOURNAL =. 2018 , NUMBER =. doi:10.1214/17-AOS1558 , URL =

work page doi:10.1214/17-aos1558 2018
[45]

Electron

Rio, Emmanuel , TITLE =. Electron. Commun. Probab. , FJOURNAL =. 2017 , PAGES =. doi:10.1214/17-ECP57 , URL =

work page doi:10.1214/17-ecp57 2017
[46]

Edwards, D. A. , TITLE =. Expo. Math. , FJOURNAL =. 2011 , NUMBER =. doi:10.1016/j.exmath.2011.06.005 , URL =

work page doi:10.1016/j.exmath.2011.06.005 2011
[47]

Computational Statistics & Data Analysis , volume=

Numerical implementation of the QuEST function , author=. Computational Statistics & Data Analysis , volume=. 2017 , publisher=

2017
[48]

Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

Maximum likelihood estimation for linear Gaussian covariance models , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2017 , publisher=

2017
[49]

2005 , PAGES =

Belinschi, Serban Teodor , TITLE =. 2005 , PAGES =

2005
[50]

Journal d'Analyse Math

A new approach to subordination results in free probability , author=. Journal d'Analyse Math. 2007 , publisher=

2007
[51]

2002 , PAGES =

Hatcher, Allen , TITLE =. 2002 , PAGES =

2002
[52]

, TITLE =

Kosorok, Michael R. , TITLE =. 2008 , PAGES =. doi:10.1007/978-0-387-74978-5 , URL =

work page doi:10.1007/978-0-387-74978-5 2008
[53]

Springer, New York

van der Vaart, Aad W. and Wellner, Jon A. , TITLE =. 1996 , PAGES =. doi:10.1007/978-1-4757-2545-2 , URL =

work page doi:10.1007/978-1-4757-2545-2 1996
[54]

Electron

Yang, Fan , TITLE =. Electron. J. Probab. , FJOURNAL =. 2019 , PAGES =. doi:10.1214/19-ejp381 , URL =

work page doi:10.1214/19-ejp381 2019
[55]

von Sydow, Bj\"orn , TITLE =. Numer. Math. , FJOURNAL =. 1977/78 , NUMBER =. doi:10.1007/BF01389313 , URL =

work page doi:10.1007/bf01389313 1977
[56]

Ledoit, Olivier and Wolf, Michael , TITLE =. Comput. Statist. Data Anal. , FJOURNAL =. 2017 , PAGES =. doi:10.1016/j.csda.2017.06.004 , URL =

work page doi:10.1016/j.csda.2017.06.004 2017
[57]

2010 , publisher=

Spectral analysis of large dimensional random matrices , author=. 2010 , publisher=

2010
[58]

Mathematische Zeitschrift , volume=

Processes with free increments , author=. Mathematische Zeitschrift , volume=. 1998 , publisher=

1998
[59]

Ding, Xiucai and Yang, Fan , TITLE =. Ann. Statist. , FJOURNAL =. 2021 , NUMBER =. doi:10.1214/20-aos1995 , URL =

work page doi:10.1214/20-aos1995 2021
[60]

http://yann.lecun.com/exdb/mnist/ , year=

The MNIST database of handwritten digits , author=. http://yann.lecun.com/exdb/mnist/ , year=
[61]

2016 , PAGES =

Gin\'e, Evarist and Nickl, Richard , TITLE =. 2016 , PAGES =. doi:10.1017/CBO9781107337862 , URL =

work page doi:10.1017/cbo9781107337862 2016
[62]

arXiv preprint arXiv:2406.05811 , year=

CLT for Generalized Linear Spectral Statistics of High-dimensional Sample Covariance Matrices and Applications , author=. arXiv preprint arXiv:2406.05811 , year=

work page arXiv
[63]

Ledoit, Olivier and P\'ech\'e, Sandrine , TITLE =. Probab. Theory Related Fields , FJOURNAL =. 2011 , NUMBER =. doi:10.1007/s00440-010-0298-3 , URL =

work page doi:10.1007/s00440-010-0298-3 2011
[64]

Raj and Mingo, James A

Rao, N. Raj and Mingo, James A. and Speicher, Roland and Edelman, Alan , TITLE =. Ann. Statist. , FJOURNAL =. 2008 , NUMBER =. doi:10.1214/07-AOS583 , URL =

work page doi:10.1214/07-aos583 2008
[65]

Tony and Liang, Tengyuan and Zhou, Harrison H

Cai, T. Tony and Liang, Tengyuan and Zhou, Harrison H. , TITLE =. J. Multivariate Anal. , FJOURNAL =. 2015 , PAGES =. doi:10.1016/j.jmva.2015.02.003 , URL =

work page doi:10.1016/j.jmva.2015.02.003 2015
[66]

Li, Zeng and Lam, Clifford and Yao, Jianfeng and Yao, Qiwei , TITLE =. Ann. Statist. , FJOURNAL =. 2019 , NUMBER =. doi:10.1214/18-AOS1782 , URL =

work page doi:10.1214/18-aos1782 2019
[67]

Li, Jun and Chen, Song Xi , TITLE =. Ann. Statist. , FJOURNAL =. 2012 , NUMBER =. doi:10.1214/12-AOS993 , URL =

work page doi:10.1214/12-aos993 2012
[68]

Jiang, Tiefeng and Yang, Fan , TITLE =. Ann. Statist. , FJOURNAL =. 2013 , NUMBER =. doi:10.1214/13-AOS1134 , URL =

work page doi:10.1214/13-aos1134 2013
[69]

Liu, Zhijun and Hu, Jiang and Bai, Zhidong and Song, Haiyan , TITLE =. Ann. Statist. , FJOURNAL =. 2023 , NUMBER =. doi:10.1214/23-aos2333 , URL =

work page doi:10.1214/23-aos2333 2023
[70]

2012 , publisher=

Introduction to non-commutative probability , author=. 2012 , publisher=

2012
[71]

and Guionnet, Alice and Zeitouni, Ofer , TITLE =

Anderson, Greg W. and Guionnet, Alice and Zeitouni, Ofer , TITLE =. 2010 , PAGES =

2010
[72]

and Zeitouni, Ofer , TITLE =

Anderson, Greg W. and Zeitouni, Ofer , TITLE =. Ann. Statist. , FJOURNAL =. 2008 , NUMBER =

2008
[73]

1999 , PAGES =

B\"ottcher, Albrecht and Silbermann, Bernd , TITLE =. 1999 , PAGES =

1999
[74]

, TITLE =

Brillinger, David R. , TITLE =. 1981 , PAGES =

1981
[75]

, TITLE =

Brillinger, David R. , TITLE =. 2001 , PAGES =

2001
[76]

Hannan, E. J. , TITLE =. 1970 , PAGES =

1970
[77]

2005 , PAGES =

L\"utkepohl, Helmut , TITLE =. 2005 , PAGES =

2005
[78]

1991 , PAGES =

L\"utkepohl, Helmut , TITLE =. 1991 , PAGES =

1991
[79]

2019 , publisher=

Multivariate time series analysis and applications , author=. 2019 , publisher=

2019
[80]

and Davis, Richard A

Brockwell, Peter J. and Davis, Richard A. , TITLE =. 2006 , PAGES =

2006

Showing first 80 references.