A0185
Title: Robust change point procedures for functional data
Authors: Alexander Duerre - TU Dortmund (Germany) [presenting]
Abstract: If functional data is observed over time, one is often confronted with the question whether its underlying distribution changes at one or several time points. The usual change point procedures have a linear structure. It is noticed for one-dimensional time series that such tests are heavily influenced by outliers which can either mask a structural change or pretend a change point. Furthermore they behave poorly under heavy tailed distributions. Therefore, robust alternatives for change point detection of functional data are presented. We apply these methods to brain activity data and respiratory air data.
