Kernel depth funcions for functional data

Abstract

In the last years the concept of data depth has been increasingly used in Statistics as a center-outward ordering of sample points in multivariate data sets. Recently data depth has been extended to functional data. In this paper we propose new intrinsic functional data depths based on the representation of functional data on Reproducing Kernel Hilbert Spaces, and test its performance against a number of well known alternatives in the problem of functional outlier detection.