32
3

Estimation of a regular conditional functional by conditional U-statistics regression

Abstract

U-statistics constitute a large class of estimators, generalizing the empirical mean of a random variable XX to sums over every kk-tuple of distinct observations of XX. They may be used to estimate a regular functional θ(PX)\theta(P_{X}) of the law of XX. When a vector of covariates ZZ is available, a conditional U-statistic may describe the effect of zz on the conditional law of XX given Z=zZ=z, by estimating a regular conditional functional θ(PXZ=)\theta(P_{X|Z=\cdot}). We prove concentration inequalities for conditional U-statistics. Assuming a parametric model of the conditional functional of interest, we propose a regression-type estimator based on conditional U-statistics. Its theoretical properties are derived, first in a non-asymptotic framework and then in two different asymptotic regimes. Some examples are given to illustrate our methods.

View on arXiv
Comments on this paper