19
2

Adaptive test for large covariance matrices with missing observations

Abstract

We observe nn independent pp-dimensional Gaussian vectors with missing coordinates, that is each value (which is assumed standardized) is observed with probability a>0a>0. We investigate the problem of minimax nonparametric testing that the high-dimensional covariance matrix Σ\Sigma of the underlying Gaussian distribution is the identity matrix, using these partially observed vectors. Here, nn and pp tend to infinity and a>0a>0 tends to 0, asymptotically. We assume that Σ\Sigma belongs to a Sobolev-type ellipsoid with parameter α>0\alpha >0. When α\alpha is known, we give asymptotically minimax consistent test procedure and find the minimax separation rates φ~n,p=(a2np)2α4α+1\tilde \varphi_{n,p}= (a^2n \sqrt{p})^{- \frac{2 \alpha}{4 \alpha +1}}, under some additional constraints on n,pn,\, p and aa. We show that, in the particular case of Toeplitz covariance matrices,the minimax separation rates are faster, ϕ~n,p=(a2np)2α4α+1\tilde \phi_{n,p}= (a^2n p)^{- \frac{2 \alpha}{4 \alpha +1}}. We note how the "missingness" parameter aa deteriorates the rates with respect to the case of fully observed vectors (a=1a=1). We also propose adaptive test procedures, that is free of the parameter α\alpha in some interval, and show that the loss of rate is (lnln(a2np))α/(4α+1)(\ln \ln (a^2 n\sqrt{p}))^{\alpha/(4 \alpha +1)} and (lnln(a2np))α/(4α+1)(\ln \ln (a^2 n p))^{\alpha/(4 \alpha +1)} for Toeplitz covariance matrices, respectively.

View on arXiv
Comments on this paper