My question is about time window for PAC (theta, 4-8 - gamma, 65-100) calculation (I am not talking about tPAC).
I have 2 conditions: each condition is 5 minutes long continuous resting EEG recording.
I wonder which option would be reasonable:
separate 5 minutes data into 60 seconds segments for each condition, run PAC on 60 seconds segments and average them per condition.
run PAC on 5 minutes long data for each condition. I wonder if in this case continuous data is still segmented into segments for PAC calculation in Brainstorm.
In theory, if there is a coupling between theta and gamma, both cases should lead to an exact same value. However, since in reality we may not have an ideal coupling and the phase of coupling may change over time, the result won't be the same.
When you calculate it for each 60 seconds you get a coupling strength and a coupling phase for all frequency pairs. If you only average the strength over these 5 parts (as we usually do), most probably you'll get a stronger value compared to when you calculate it from the whole interval (small shifts in phase will not be considered in averaging).
As far as I know, there is no guideline about which approach is the correct one. You can go with either options, but the common way of doing it is using the whole interval for PAC. You can also try tPAC with 60 seconds sliding windows.
(No, PAC with MVL method from Canolty does not break data into smaller chunks. )