Functional Outlier Detection through Probabilistic Modelling

Functional data consist in the most typical case of one-dimensional curves that represent the evolution of some physical parameter of interest with time. However, the analysis of this kind of objects is far from being simple, and the possibility of treating contaminated data is a classical problem that can arise in this framework as frequently as in the multivariate one. This justifies the development a new functional outlier detection technique based on functional measures capable of capturing the the outlyingness in the magnitude and shape sense that is presented in this paper.