Investigating Dependence in Multivariate Time Series

• Activity Number: 56
• Type: Topic Contributed
• Date/Time: Sunday, August 1, 2010 : 4:00 PM to 5:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Hernando Ombao*+ and Mark Fiecas and Cristina Gorrostieta
• Companies: Brown University and Brown University and Brown University
• Keywords: Multivariate signals ; Partial coherence ; Mutual information ; Spectral analysis ; multi-scale transforms

Abstract

We develop multi-scale methods and models for characterizing, estimating and comparing dependence in multivariate non-stationary time series. This is motivated by a study on pediatric bipolar patients (led by D. Dickstein, Brown University) where one goal is to study the circuit involving the amygdala, accumbens, temporal and dorsolateral pre-frontal cortex - which are engaged in emotion regulation, decision making and attention. In this talk, we shall characterize dependence using a variety of multi-scale measures such as the evolutionary partial coherence, partial correlation and mutual information. We shall develop these measures under models of non-stationary time series that use localized transforms. We shall develop generalized shrinkage methods for estimation and randomization tests for comparing dependence between bipolar patients and healthy controls.