Title: Functional shape outliers as contamination in complexity mixtures Authors:  Enea Bongiorno - Universita del Piemonte Orientale (Italy) [presenting]
Aldo Goia - Universita' del Piemonte Orientale (Italy)
Kwo Lik Lax Chan - Universita degli Studi del Piemonte Orientale (Italy)
Abstract: Shape outliers are treated as contamination elements and part of a high-complexity component within an appropriate mixture model of functional data. The aim is threefold. First, we define the notion of complexity based on the concept of small ball probability. Second, we theoretically introduce the idea of a complexity mixture and analyze its implications on small ball probabilities. Third, we propose an algorithm to decompose a complexity mixture into its constituent components, thereby implicitly identifying potential contamination in a functional dataset. The effectiveness of the proposed methodology is demonstrated through an application.