A0636
Title: FKWC: Robust covariance changepoint detection for multivariate functional data
Authors: Kelly Ramsay - York University (Canada) [presenting]
Shojaeddin Chenouri - University of Waterloo (Canada)
Abstract: A novel changepoint detection method is proposed, the Functional Kruskal-Wallis for Covariance (FKWC) procedure, which leverages rank statistics and multivariate functional data depth. The FKWC procedure robustly identifies changepoints in the variability of sequences of independent multivariate functional data. It supports testing for at most one changepoint, detecting epidemic changes, or estimating both the number and locations of multiple unknown changepoints. After considering various theoretical and empirical results, applications of the FKWC procedure to fMRI scans and functional financial time series are showcased.
