Weak convergence of empirical Wasserstein type distances

arXiv1911.02389MaRDI QIDQ6328657FDOQ6328657

Publication date: 6 November 2019

Abstract: We estimate contrasts

i n t_{0}^{1} h o (F^{- 1} (u) - G^{- 1} (u)) d u

between two continuous distributions

F

and

G

on

m a t h b b R

such that the set

F = G

is a finite union of intervals, possibly empty or

m a t h b b R

. The non-negative convex cost function

h o

is not necessarily symmetric and the sample may come from any joint distribution

H

on

m a t h b b R^{2}

with marginals

F

and

G

having light enough tails with respect to

h o

. The rates of weak convergence and the limiting distributions are derived in a wide class of situations including the classical Wasserstein distances

W_{1}

and

W_{2}

. The new phenomenon we describe in the case

F = G

involves the behavior of

h o

near

0

, which we assume to be regularly varying with index ranging from

1

to

2

and to satisfy a key relation with the behavior of

h o

near

i n f t y

through the common tails. Rates are then also regularly varying with powers ranging from

1 / 2

to

1

also affecting the limiting distribution, in addition to

H

.

This page was built for publication: Weak convergence of empirical Wasserstein type distances

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6328657)