[FieldTrip] Cross spectral density versus coherence

[Jord Vink]

 Hi all,

I am working with functional connectivity between different (EEG) signals.
I'm calculating the connectivity strength between different brain regions
and correlate the connectivity strength to a certain outcome measure. Among
others, I'm using the cross spectral density (CSD) and the coherence. The
CSD is not generally used as a measure of connectivity but I don't
completely understand why not. Especially since the CSD correlates well
with my outcome measure, while the coherence doesn't. I'm trying to
undertand whether there is a fundamental reason why I can't use the CSD,
and why I should use the coherence.

So I'm trying to understand why people use coherence (the normalized
version of the cross spectral density) as a measure of connectivity instead
of cross spectral density (CSD). Why do you have to normalize the CSD with
the autospectral densities to determine connectivity when the CSD by itself
is a measure of signal similarity? Is there a fundamental reason to
normalize the CSD?

Jord
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20180204/25fdffd4/attachment-0001.html>