Estimation of smooth functionals in normal models: bias reduction and asymptotic efficiency

DOI10.1214/20-AOS2047zbMATH Open1486.62163arXiv1912.08877OpenAlexW3213029249MaRDI QIDQ2054520FDOQ2054520

Authors: M. Zhilova, Vladimir Koltchinskii

Publication date: 3 December 2021

Published in: The Annals of Statistics (Search for Journal in Brave)

Abstract: Let

X_{1}, d o t s, X_{n}

be i.i.d. random variables sampled from a normal distribution

N (m u, S i g m a)

in

{m a t h b b R}^{d}

with unknown parameter

h e t a = (m u, S i g m a) i n T h e t a := {m a t h b b R}^{d} i m e s {m a t h c a l C}_{+}^{d},

where

{m a t h c a l C}_{+}^{d}

is the cone of positively definite covariance operators in

{m a t h b b R}^{d} .

Given a smooth functional

f : T h e t a m a p s t o {m a t h b b R}^{1},

the goal is to estimate

f (h e t a)

based on

X_{1}, d o t s, X_{n} .

Let

Theta(a;d):={mathbb R}^d imes Bigl{Sigmain {mathcal C}_+^d: sigma(Sigma)subset [1/a, a]Bigr}, ageq 1,

where

s i g m a (S i g m a)

is the spectrum of covariance

S i g m a .

Let

h a t h e t a := (h a t m u, h a t S i g m a),

where

h a t m u

is the sample mean and

h a t S i g m a

is the sample covariance, based on the observations

X_{1}, d o t s, X_{n} .

For an arbitrary functional

f i n C^{s} (T h e t a),

s = k + 1 + h o, k g e q 0, h o i n (0, 1],

we define a functional

f_{k} : T h e t a m a p s t o m a t h b b R

such that �egin{align*} & sup_{ hetain Theta(a;d)}|f_k(hat heta)-f( heta)|_{L_2({mathbb P}_{ heta})} lesssim_{s, �eta} |f|_{C^{s}(Theta)} �iggr[�iggl(frac{a}{sqrt{n}} �igvee a^{�eta s}�iggl(sqrt{frac{d}{n}}�iggr)^{s} �iggr)wedge 1�iggr], end{align*} where

for