[FieldTrip] Question about Implementation of Partial Directed Coherence

[Makhoul, Ghassan Sebastian Peter]

Good evening,

I am analyzing intracranial EEG recordings to understand epilepsy. I want to use partial directed coherence (PDC) to characterize relationships between channels in the intracranial EEG. I am new to PDC and am also brushing up on my signal processing so I could use a bit of help.

I used the fieldtrip implementation of PDC and have found that connectivity across all frequency bands of interest seems highly similar. This is odd to me, as I would at least expect long range connectivity to bin in the low frequency range with high frequencies representing local interactions. Nonetheless, all connectivity seems similar across the full range of frequency bands. I worry I missed something essential. One thought I had was in the specification of the model order for the MVAR.

I use a model order of 10 with a sampling frequency of 500Hz. If I am to understand correctly, does that mean that my MVAR model only looks at .02s (10/500) of information when determining connectivity? If that's the case, then will this not severely limit its ability to pick up low frequency information? I think the root of my confusion is that I don't understand the mechanics of the frequency decomposition.

Any help would be greatly appreciated!

Thanks,
Ghassan Makhoul

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20241018/006a58ea/attachment.htm>