Hello!
I am working on 12-channel intracerebral EEG data acquired in the sleeping state. I don’t have any epochs/events in my data. They are about a minute long. I want to obtain a 12 by 12 (symmetrical) matrix that has the coherence value of each pair of channels in each entry. I would like to have this matrix for different frequency bands, such as alpha, beta, gamma, etc.
My first confusion is our assumptions while computing the PSD and CSD. If I am not mistaken, we assume the signal is wide sense stationary while computing PSD. However, my one-minute-long data is indeed not stationary. Is it still alright to compute PSD or CSD over this one-minute-long time series data?
When the first confusion is cleared, my second question becomes: “Which MNE-connectivity function to compute this coherence?” As far as I understand, “spectral_connectivity_time” computes the coherence over time of one epoch. This is not what I wanted. The other option seems to be " spectral_connectivity_epochs." However, this function computes the average of PSDs/CSDs over epochs, and it is also mentioned in its document that using only one epoch would yield errorful results. Other than these two, I could not find any related functions.
Are there any functions in MNE-connectivity that I can use on my one-minute-long 12-channel data to compute interchannel coherence for different frequency bands?
I am relatively new to the field. Any help, clarification, or correction is appreciated 
