We introduce a class of sample covariance matrices which subsumes and generalizes several previous models. The associated population covariance matrix is assumed to differ from the identity by a matrix of bounded rank. All quantities except the rank of may depend on in an arbitrary fashion. We investigate the principal components, i.e.\ the top eigenvalues and eigenvectors, of . We derive precise large deviation estimates on the generalized components of the outlier and non-outlier eigenvectors . Our results also hold near the so-called BBP transition, where outliers are created or annihilated, and for degenerate or near-degenerate outliers. We believe the obtained rates of convergence to be optimal. In addition, we derive the asymptotic distribution of the generalized components of the non-outlier eigenvectors. A novel observation arising from our results is that, unlike the eigenvalues, the eigenvectors of the principal components contain information about the \emph{subcritical} spikes of . The proofs use several results on the eigenvalues and eigenvectors of the uncorrelated matrix , satisfying , as input: the isotropic local Marchenko-Pastur law established in [9], level repulsion, and quantum unique ergodicity of the eigenvectors. The latter is a special case of a new universality result for the joint eigenvalue-eigenvector distribution.
View on arXiv