Beyond Linear Dynamic Functional Connectivity: A Vine Copula Change Point Model

To estimate dynamic functional connectivity for magnetic resonance imaging (fMRI) data, two approaches have dominated: sliding window and change point methods. While computationally feasible, the approach has several limitations. In addition, existing methods assume a Gaussian distribution linear dependencies between fMRI time series. this work, we introduce new methodology called Vine Copula Change Point (VCCP) to points in network structure brain regions. It uses vine copulas, various state-of-the-art segmentation identify multiple points, likelihood ratio test or stationary bootstrap inference. The copulas allow forms of dependence regions including tail, symmetric asymmetric dependence, which not been explored before analysis neuroimaging data. We apply VCCP simulation datasets datasets: reading task an anxiety inducing experiment. particular, former dataset, illustrate complexity textual changes during Chapter 9 Harry Potter Sorcerer’s Stone find that across subjects are related more than one type attributes. Further, graphs created by indicate importance working beyond Gaussianity dependence. Finally, R package vccp implementing from article is available CRAN. Supplementary Materials online.