46
109

Optimal hypothesis testing for high dimensional covariance matrices

Abstract

This paper considers testing a covariance matrix Σ\Sigma in the high dimensional setting where the dimension pp can be comparable or much larger than the sample size nn. The problem of testing the hypothesis H0:Σ=Σ0H_0:\Sigma=\Sigma_0 for a given covariance matrix Σ0\Sigma_0 is studied from a minimax point of view. We first characterize the boundary that separates the testable region from the non-testable region by the Frobenius norm when the ratio between the dimension pp over the sample size nn is bounded. A test based on a UU-statistic is introduced and is shown to be rate optimal over this asymptotic regime. Furthermore, it is shown that the power of this test uniformly dominates that of the corrected likelihood ratio test (CLRT) over the entire asymptotic regime under which the CLRT is applicable. The power of the UU-statistic based test is also analyzed when p/np/n is unbounded.

View on arXiv
Comments on this paper