Estimation of smooth functionals of covariance operators: jackknife bias reduction and bounds in terms of effective rank
share
Let E be a separable Banach space and let X, X_1,…, X_n, … be i.i.d. Gaussian random variables taking values in E with mean zero and unknown covariance operator Σ: E^∗↦ E. The complexity of estimation of Σ based on observations X_1,…, X_n is naturally characterized by the so called effective rank of Σ: r(Σ):= 𝔼_ΣX^2/Σ, where Σ is the operator norm of Σ. Given a smooth real valued functional f defined on the space L(E^∗,E) of symmetric linear operators from E^∗ into E (equipped with the operator norm), our goal is to study the problem of estimation of f(Σ) based on X_1,…, X_n. The estimators of f(Σ) based on jackknife type bias reduction are considered and the dependence of their Orlicz norm error rates on effective rank r(Σ), the sample size n and the degree of Hölder smoothness s of functional f are studied. In particular, it is shown that, if r(Σ)≲ n^α for some α∈ (0,1) and s≥1/1-α, then the classical √(n)-rate is attainable and, if s> 1/1-α, then asymptotic normality and asymptotic efficiency of the resulting estimators hold. Previously, the results of this type (for different estimators) were obtained only in the case of finite dimensional Euclidean space E=ℝ^d and for covariance operators Σ whose spectrum is bounded away from zero (in which case, r(Σ)≍ d).
READ FULL TEXT