Double Threshold GARCH Model with Applications to EEG Data

• Activity Number: 232
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Sipan Aslan*+ and Hernando Ombao
• Companies: Middle East Technical University and University of California, Irvine
• Keywords: Threshold Nonlinearity ; Double Threshold ; GARCH ; EEG ; Switching Regimes ; Time Series

Abstract

Most dynamical system, particularly high dimensional systems with the complex structures such as brain electrophysiological signals do not exhibit a uniform type of behavior over a time. In order to analyze time-varying behavior of the observed system one can use the notion of the changing regimes where the each regime describes the certain characteristics of the process. This study presents the new threshold time series modelling procedure for brain signals by utilizing the regimes switching in the mean and the conditional variance which are achieved by implementing threshold nonlinearity and the threshold GARCH model, respectively. Such robustification on determining the threshold variable, models orders and selection criteria are proposed. The model will be investigated via extensive numerical experiments and demonstrated on EEG signals recorded during an epileptic seizure.