B1686
Title: Graph inference from multivariate time series with long-range dependence
Authors: Irene Gannaz - INSA Lyon (France) [presenting]
Abstract: Brain organization, or functional connectivity, is characterized. This can be estimated by the correlation between signals measuring brain activity. Time series have inhomogeneous long-range dependence properties. An estimation procedure in a semi-parametric framework is proposed, based on a Whittle approximation of the wavelet representation. Asymptotic normality is established for the long-range dependence parameters and the long-range correlations. It is used to show that long-range dependence is associated with brain activity. A graphical representation of functional connectivity is inferred by significance tests on the correlations.
