Robust functional principal component analysis

Abstract

When dealing with multivariate data robust PCA, like classical PCA,searches for directions with maximal dispersion of the data projected on it. Insteadof using the variance as a measure of dispersion, a robust scale estimator sn may beused in the maximization problem. In this paper, we review some of the proposedapproaches to robust functional PCA including one which adapts the projection pursuitapproach to the functional data setting.