Let be i.i.d. Gaussian random variables with zero mean and covariance operator taking values in a separable Hilbert space Let {\bf r}(\Sigma):=\frac{{\rm tr}(\Sigma)}{\|\Sigma\|_{\infty}} be the effective rank of being the trace of and being its operator norm. Let \hat \Sigma_n:=n^{-1}\sum_{j=1}^n (X_j\otimes X_j) be the sample (empirical) covariance operator based on The paper deals with a problem of estimation of spectral projectors of the covariance operator by their empirical counterparts, the spectral projectors of (empirical spectral projectors). The focus is on the problems where both the sample size and the effective rank are large. This framework includes and generalizes well known high-dimensional spiked covariance models. Given a spectral projector corresponding to an eigenvalue of covariance operator and its empirical counterpart we derive sharp concentration bounds for bilinear forms of empirical spectral projector in terms of sample size and effective dimension Building upon these concentration bounds, we prove the asymptotic normality of bilinear forms of random operators under the assumptions that and In a special case of eigenvalues of multiplicity one, these results are rephrased as concentration bounds and asymptotic normality for linear forms of empirical eigenvectors. Other results include bounds on the bias and a method of bias reduction as well as a discussion of possible applications to statistical inference in high-dimensional principal component analysis.
View on arXiv