We present visualization and statistical exploratory approaches to modeling dependence between components of high-dimensional time series in designed experiments. Our goal is to study the impact of an event (e.g., stroke) on the dependence structure in the short-term and long-term between brain regions. Our approach is to decompose the activity of populations of neurons into their various oscillatory components. This decomposition allows the examination and computation of cross-correlation between different oscillatory activities. Furthermore, the process of computing the cross-correlation is carried through a condensed and padded tensor multiplication, providing information about the time shift and unbiased co-variance values. The proposed approach provides both scalability and efficiency to perform cross-correlation computations on large datasets.
This is a joint work with H. Ombao (KAUST) and R. Frostig (UC Irvine).
|