Edge universality of correlation matrices

DOI10.1214/12-AOS1022MaRDI QIDQ693745zbMATH OpenFDO

Authors Natesh S. Pillai, Jun Yin

Publication date 10 December 2012

Published in The Annals of Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1112.2381, https://projecteuclid.org/euclid.aos/1349196390

zbMATH Keywords

correlation matrixcovariance matrixMarcenko-Pastur lawTracy-Widom lawuniversality

Mathematics Subject Classification ID

Random matrices (algebraic aspects) (15B52) Measures of association (correlation, canonical correlation, etc.) (62H20) Analysis of variance and covariance (ANOVA) (62J10) Disordered systems (random Ising models, random Schrödinger operators, etc.) in equilibrium statistical mechanics (82B44)

Abstract: Let $widetilde{X}_{MimesN}$ be a rectangular data matrix with independent real-valued entries $[widetilde{x}_{ij}]$ satisfying $mathbbEwidetilde{x}_{ij}=0$ and $mathbbEwidetilde{x}_{ij}^2=frac1M$, $N,Moinfty$. These entries have a subexponential decay at the tails. We will be working in the regime $N/M=d_N,li{m}_{Noinfty}{d}_{N}eq0,1,infty$. In this paper we prove the edge universality of correlation matrices $X^{dagger}X$, where the rectangular matrix $X$ (called the standardized matrix) is obtained by normalizing each column of the data matrix $widetildeX$ by its Euclidean norm. Our main result states that asymptotically the $k$-point ($kgeq1$) correlation functions of the extreme eigenvalues (at both edges of the spectrum) of the correlation matrix $X^{dagger}X$ converge to those of the Gaussian correlation matrix, that is, Tracy-Widom law, and, thus, in particular, the largest and the smallest eigenvalues of $X^{dagger}X$ after appropriate centering and rescaling converge to the Tracy-Widom distribution. The asymptotic distribution of extreme eigenvalues of the Gaussian correlation matrix has been worked out only recently. As a corollary of the main result in this paper, we also obtain that the extreme eigenvalues of Gaussian correlation matrices are asymptotically distributed according to the Tracy-Widom law. The proof is based on the comparison of Green functions, but the key obstacle to be surmounted is the strong dependence of the entries of the correlation matrix. We achieve this via a novel argument which involves comparing the moments of product of the entries of the standardized data matrix to those of the raw data matrix. Our proof strategy may be extended for proving the edge universality of other random matrix ensembles with dependent entries and hence is of independent interest.

Review text

The edge universality of correlation matrices is proved. The term universality refers to the phenomenon that the asymptotic distributions of various functionals of covariance/correlation matrices are identical to those Gaussian covariance/correlation matrices. The here considered data matrix \(\tilde X = (\tilde x_{ij})\) is an \(M \times N\) matrix with independent centered real-valued entries. The entries in each column \(j\) are assumed to be identically distributed, i.~e., \(\tilde x_{ij} = M^{-1/2}q_{ij}\), \({\mathbb E}q_{ij}=0\) and \({\mathbb E}q_{ij}^2= \sigma_j^2\) for \(1 < i < M\). These entries \(q_{ij}\) have a subexponential decay, i. e., there exists a constant \(\vartheta > 0\) such that for \(u >1\) we have \({\mathbb P}(|q_{ij}| > u \sigma_j) \leq \vartheta^{-1} \text{exp}(-u^{ \vartheta})\). It is assumed that \(\lim_{N \to \infty} d_N \not= 0, 1, \infty\) is valid for \(d_N = N/M\). The main result of the paper says that the \(k\)-point correlation functions of the extreme eigenvalues of the correlation matrix \(X^{\dagger}X\) converge to those of the Gaussian correlation matrix, i. e., that the asymptotic distribution of the largest or smallest eigenvalue of the correlation matrix is given by the Tracy-Widom law. The proof is based on the comparison of Green's functions but doing this it is necessary to overcome the strong dependence of the entries of the correlation matrix. It is achieved by the new argument which includes comparing the moments of the product of the entries of the standardized data matrix to those of the raw data matrix. As a special case of the main result the asymptotic distribution of the largest or smallest (appropriately rescaled) eigenvalue of the Gaussian correlation matrix is obtained.

Recommendations

• Edge universality of correlated Gaussians
• Universality of covariance matrices
• Edge universality of separable covariance matrices
• Correlated random matrices: band rigidity and edge universality
• Universality of random matrices with correlated entries

Cites work

• A generalization of the Lindeberg principle
• A note on universality of the distribution of the largest eigenvalues in certain sample covariance matrices
• A universality result for the smallest eigenvalues of certain sample covariance matrices
• Bulk universality for generalized Wigner matrices
• Bulk universality for Wigner matrices
• Forward-backward SDEs with random terminal time and applications to pricing special European-type options for a large investor
• scientific article; zbMATH DE number 5204610 (Why is no real title available?)
• scientific article; zbMATH DE number 3244317 (Why is no real title available?)
• Local semicircle law and complete delocalization for Wigner random matrices
• Multivariate analysis and Jacobi ensembles: largest eigenvalue, Tracy-Widom limits and rates of convergence
• On the distribution of the largest eigenvalue in principal components analysis
• Random covariance matrices: universality of local statistics of eigenvalues
• Random matrices: Universality of local eigenvalue statistics up to the edge
• Rigidity of eigenvalues of generalized Wigner matrices
• Semicircle law on short scales and delocalization of eigenvectors for Wigner random matrices
• Spectral statistics of Erdős-Rényi graphs. I: Local semicircle law
• The isotropic semicircle law and deformation of Wigner matrices
• The limiting distributions of eigenvalues of sample correlation matrices
• The local relaxation flow approach to universality of the local statistics for random matrices
• Tracy-Widom law for the extreme eigenvalues of sample correlation matrices
• Universality for certain Hermitian Wigner matrices under weak moment conditions
• Universality for generalized Wigner matrices with Bernoulli distribution
• Universality of covariance matrices
• Universality of random matrices and local relaxation flow
• Universality results for the largest eigenvalues of some sample covariance matrix ensembles
• Wegner Estimate and Level Repulsion for Wigner Random Matrices

Cited in

(31)

• Recent developments in high dimensional covariance estimation and its related issues, a review
• A necessary and sufficient condition for edge universality at the largest singular values of covariance matrices
• Properties of eigenvalues and eigenvectors of large-dimensional sample correlation matrices
• Limiting distributions for eigenvalues of sample correlation matrices from heavy-tailed populations
• Detection of hubs in complex networks by the Laplacian matrix
• Correlated random matrices: band rigidity and edge universality
• Tracy-Widom limit for Kendall's tau
• Universality for the largest eigenvalue of sample covariance matrices with general population
• A necessary and sufficient condition for edge universality of Wigner matrices
• Random matrix theory in statistics: a review
• Universality of covariance matrices
• Edge universality of correlated Gaussians
• Central limit theorem of linear spectral statistics of high-dimensional sample correlation matrices
• Tracy-Widom law for the extreme eigenvalues of sample correlation matrices
• Separation of the largest eigenvalues in eigenanalysis of genotype data from discrete subpopulations
• Asymptotics of eigenstructure of sample correlation matrices for high-dimensional spiked models
• Singular vector distribution of sample covariance matrices
• A test for the complete independence of high-dimensional random vectors
• Asymptotic independence of the extreme eigenvalues of Gaussian unitary ensemble
• Considering Horn's parallel analysis from a random matrix theory point of view
• On eigenvalues of a high-dimensional Kendall's rank correlation matrix with dependence
• Quantitative universality for the largest eigenvalue of sample covariance matrices
• Norm convergence rate for multivariate quadratic polynomials of Wigner matrices
• Sampling without replacement from a high-dimensional finite population
• Numerical algorithms for the asymptotic mean and variance of linear spectral statistics of high-dimensional correlation matrices
• An adaptive test of the independence of high-dimensional data based on Kendall rank correlation coefficient
• Testing high-dimensional covariance structures using double-normalized observations
• Extreme eigenvalues of log-concave ensemble
• Spectral properties of high-dimensional rescaled sample correlation matrices
• Central limit theorem for linear spectral statistics of large dimensional Kendall's rank correlation matrices and its applications
• Spectral statistics of large dimensional Spearman's rank correlation matrix and its application

This page was built for publication: Edge universality of correlation matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q693745)