Sample canonical correlation coefficients of high-dimensional random vectors: Local law and Tracy–Widom limit

DOI10.1142/S2010326322500071zbMATH Open1526.60011arXiv2002.09643MaRDI QIDQ6063733FDOQ6063733

Authors:

Publication date: 8 November 2023

Published in: Random Matrices: Theory and Applications (Search for Journal in Brave)

Abstract: Consider two random vectors $mathbf{C}_1^{1/2}mathbfxinmathbb{R}^p$ and $mathbf{C}_2^{1/2}mathbfyinmathbb{R}^q$, where the entries of $mathbfx$ and $mathbfy$ are i.i.d. random variables with mean zero and variance one, and $mathbf{C}_1$ and $mathbf{C}_2$ are $pimesp$ and $qimesq$ deterministic population covariance matrices. With $n$ independent samples of $(mathbf{C}_1^{1/2}mathbfx,mathbf{C}_2^{1/2}mathbfy)$, we study the sample correlation between these two vectors using canonical correlation analysis. We denote by $S_{xx}$ and $S_{yy}$ the sample covariance matrices for $mathbf{C}_1^{1/2}mathbfx$ and $mathbf{C}_2^{1/2}mathbfy$, respectively, and $S_{xy}$ the sample cross-covariance matrix. Then the sample canonical correlation coefficients are the square roots of the eigenvalues of the sample canonical correlation matrix $cal{C}_{XY}:=S_{xx}^{-1}{S}_{xy}{S}_{yy}^{-1}{S}_{yx}$. Under the high-dimensional setting with $p/no{c}_{1}in(0,1)$ and $q/no{c}_{2}in(0,1-c_1)$ as $noinfty$, we prove that the largest eigenvalue of $mathcal{C}_{XY}$ converges to the Tracy-Widom distribution as long as we have $li{m}_{sightarrowinfty}{s}^4[mathbbP(vert{x}_{ij}vertgeqs)+mathbbP(vert{y}_{ij}vertgeqs)]=0$. This extends the result in [16], which established the Tracy-Widom limit of the largest eigenvalue of $mathcal{C}_{XY}$ under the assumption that all moments are finite. Our proof is based on a linearization method, which reduces the problem to the study of a $(p+q+2n)imes(p+q+2n)$ random matrix $H$. In particular, we shall prove an optimal local law on its inverse $G:=H^{-1}$, i.e the resolvent. This local law is the main tool for both the proof of the Tracy-Widom law in this paper, and the study in [22,23] on the canonical correlation coefficients of high-dimensional random vectors with finite rank correlations.

Full work available at URL: https://arxiv.org/abs/2002.09643

Recommendations

• Limiting distribution of the sample canonical correlation coefficients of high-dimensional random vectors
• Sample canonical correlation coefficients of high-dimensional random vectors with finite rank correlations
• Canonical correlation coefficients of high-dimensional Gaussian vectors: finite rank case
• Tracy-Widom law for the extreme eigenvalues of sample correlation matrices
• The convergence of the empirical distribution of canonical correlation coefficients

zbMATH Keywords

canonical correlation analysislocal lawTracy-Widom law

Mathematics Subject Classification ID

Multivariate analysis (62H99) Asymptotic distribution theory in statistics (62E20) Random matrices (probabilistic aspects) (60B20)

Cites Work

• A necessary and sufficient condition for edge universality of Wigner matrices
• RELATIONS BETWEEN TWO SETS OF VARIATES
• Rigidity of eigenvalues of generalized Wigner matrices
• No eigenvalues outside the support of the limiting spectral distribution of large-dimensional sample covariance matrices
• Multivariate analysis and Jacobi ensembles: largest eigenvalue, Tracy-Widom limits and rates of convergence
• The spectrum edge of random matrix ensembles.
• High-dimensional asymptotic distributions of characteristic roots in multivariate linear models and canonical correlation analysis
• Averaging fluctuations in resolvents of random band matrices
• The local semicircle law for a general class of random matrices
• Isotropic local laws for sample covariance and generalized Wigner matrices
• Local circular law for random matrices
• The limiting empirical measure of multiple discriminant ratios
• Universality of covariance matrices
• The Tracy-Widom law for the largest eigenvalue of F type matrices
• Minimax estimation in sparse canonical correlation analysis
• Independence test for high dimensional data based on regularized canonical correlation coefficients
• Spectral statistics of Erdős-Rényi graphs. I: Local semicircle law
• Delocalization and diffusion profile for random band matrices
• The convergence of the empirical distribution of canonical correlation coefficients
• On the principal components of sample covariance matrices
• Local law for random Gram matrices
• A necessary and sufficient condition for edge universality at the largest singular values of covariance matrices
• Local circular law for the product of a deterministic matrix with a random matrix
• Sparse CCA: adaptive estimation and computational barriers
• Canonical correlation coefficients of high-dimensional Gaussian vectors: finite rank case
• Singularities of the density of states of random Gram matrices
• A unified matrix model including both CCA and F matrices in multivariate analysis: the largest eigenvalue and its applications
• Testing in high-dimensional spiked models
• Spiked separable covariance matrices and principal components
• Convergence of eigenvector empirical spectral distribution of sample covariance matrices
• Edge universality of separable covariance matrices

Cited In (6)

• Limiting distribution of the sample canonical correlation coefficients of high-dimensional random vectors
• Convergence rate to the Tracy-Widom laws for the largest eigenvalue of sample covariance matrices
• Sample canonical correlation coefficients of high-dimensional random vectors with finite rank correlations
• Spiked eigenvalues of noncentral Fisher matrix with applications
• Limiting spectral distribution of high-dimensional noncentral Fisher matrices and its analysis
• Local canonical correlation coefficients and canonical variables of groups of random variables

This page was built for publication: Sample canonical correlation coefficients of high-dimensional random vectors: Local law and Tracy–Widom limit

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6063733)