Asymptotically Efficient Estimation of Smooth Functionals of Covariance Operators

Let be a centered Gaussian random variable in a separable Hilbert space with covariance operator We study a problem of estimation of a smooth functional of based on a sample of independent observations of More specifically, we are interested in functionals of the form where is a smooth function and is a nuclear operator in We prove concentration and normal approximation bounds for plug-in estimator being the sample covariance based on These bounds show that is an asymptotically normal estimator of its expectation (rather than of parameter of interest ) with a parametric convergence rate provided that the effective rank ( being the trace and being the operator norm of ) satisfies the assumption At the same time, we show that the bias of this estimator is typically as large as (which is larger than if ). In the case when is finite-dimensional space of dimension we develop a method of bias reduction and construct an estimator of that is asymptotically normal with convergence rate Moreover, we study asymptotic properties of the risk of this estimator and prove minimax lower bounds for arbitrary estimators showing the asymptotic efficiency of in a semi-parametric sense.
View on arXiv