Canonical correlation coefficients of high-dimensional Gaussian vectors: finite rank case
arXiv:1704.02408
Abstract
Consider a Gaussian vector , consisting of two sub-vectors and with dimensions and respectively, where both and are proportional to the sample size . Denote by the population cross-covariance matrix of random vectors and , and denote by the sample counterpart. The canonical correlation coefficients between and are known as the square roots of the nonzero eigenvalues of the canonical correlation matrix . In this paper, we focus on the case that is of finite rank , i.e. there are nonzero canonical correlation coefficients, whose squares are denoted by . We study the sample counterparts of , i.e. the largest eigenvalues of the sample canonical correlation matrix , denoted by . We show that there exists a threshold , such that for each , when , converges almost surely to the right edge of the limiting spectral distribution of the sample canonical correlation matrix, denoted by . When , possesses an almost sure limit in . We also obtain the limiting distribution of 's under appropriate normalization. Specifically, possesses Gaussian type fluctuation if , and follows Tracy-Widom distribution if . Some applications of our results are also discussed.
arXiv commentThis is an extended version of the previous work arXiv:1407.7194v2. In the current work, we have included the result on the fluctuations, and the limit part has also been reorganized
References in corpus (6)
- The largest eigenvalue of rank one deformation of large Wigner matrices
- The largest eigenvalues of finite rank deformation of large Wigner matrices: convergence and nonuniversality of the fluctuations
- The largest eigenvalues of sample covariance matrices for a spiked population: diagonal case
- Sparse CCA: Adaptive Estimation and Computational Barriers
- Subordination for the sum of two random matrices
- Convergence of the empirical spectral distribution function of Beta matrices