Some asymptotic results on density estimators by wavelet projections

Abstract: Let

(X_{i})_{i g e q 1}

be an i.i.d. sample on

R R R^{d}

having density

f

. Given a real function

p h i

on

R R R^{d}

with finite variation and given an integer valued sequence

(j_{n})

, let

f n

denote the estimator of

f

by wavelet projection based on

p h i

and with multiresolution level equal to

j_{n}

. We provide exact rates of almost sure convergence to 0 of the quantity

s u p_{x i n H} m i d f n (x) - E E E (f n) (x) m i d

, when

n 2^{- d j_{n}} / l o g n a r i n f t y

and

H

is a given hypercube of

R R R^{d}

. We then show that, if

n 2^{- d j_{n}} / l o g n a r c

for a constant

c > 0

, then the quantity

s u p_{x i n H} m i d f n (x) - f m i d

almost surely fails to converge to 0.

Recommendations

Cites work

Cited in

(7)

This page was built for publication: Some asymptotic results on density estimators by wavelet projections