Coherence analysis of multichannel time series applying conditioned multivariate autoregressive spectra

Coherence analysis enables the studying of linear dependencies between multichannel time series. In the case of a multivariate autoregressive (MAR) spectrum the conventional coherence analysis can be applied. However, since we are able to decompose the MAR spectrum, there is a possibility to gain more information through coherence analysis based on conditioned spectra than with conventional methods. The authors formulate the coherence analysis based on the conditioned MAR spectra (reduced and noise conditioned spectra) by giving related definitions for partial and multiple coherences.<<ETX>>