Measuring Connectivity in Linear Multivariate Processes: Definitions, Interpretation, and Practical Analysis

摘要: This tutorial paper introduces a common framework for the evaluation of widely used frequency-domain measures coupling (coherence, partial coherence) and causality (directed coherence, directed from parametric representation linear multivariate (MV) processes. After providing comprehensive time-domain definition various forms connectivity observed in MV processes, we particularize them to autoregressive (MVAR) processes derive corresponding measures. Then, discuss theoretical interpretation these MVAR-based measures, showing that each reflects specific how this results description peculiar aspects information transfer Furthermore, issues related practical utilization on real-time series are pointed out, including MVAR model estimation significance assessment. Finally, limitations pitfalls arising mis-specification discussed, indicating possible solutions recommendations safe computation An example presented multiple EEG signals recorded during combined visuomotor task is also reported, frequency domain may help describing neurophysiological mechanisms.