[FieldTrip] some general questions about gamma-related analyses

[Tzvetan Popov]

Dear Ivo,

> The whole of the gamma range expands well beyond 30 Hz, the noise threshold, and thus the first question is whether low-pass filtering below 30 Hz is at all applied when looking for gamma coherence, power etc. ...  One reason I kept the classical filter settings at first is to be consistent across the many different analyses I apply to the same data sets, not all of which have to do with gamma, but it dawned on me that filtering out frequencies above 30 is probably not a good idea(?) for gamma-related questions?
No, do not apply 30Hz LP. You filter out your signal of interest.
> 
> Moreover, is it acceptable to look for the point of maximum coherence within that range, rather than average within the range? The data suggests that the point of maximum coherence shifts across time to higher and higher frequencies, thus I decided for the time being to report the maximum coherence (across electrodes) value within the gamma range.
One typically applies smoothening in the frequency domain when evaluating gamma (power and coherence). I suggest to study this lecture https://www.youtube.com/watch?v=dHTuzMsjVJA <https://www.youtube.com/watch?v=dHTuzMsjVJA>. In particular, at minute 25:49 the advantage of the freq smoothening is highlighted. Given that example your peak frequency will change depending on your decision how much to smooth.

Of note, channel level coherence, particularly for gamma is not recommended. A lot of other stuff is added to the mix (e.g. cardiac, neck muscles etc.). Their contribution will change over time in your experiment. At the end you will not be able to interpret the results and/or you might be fooled.
I suggest you also study this lecture https://www.youtube.com/watch?v=ZBwh0Vm4fh4 <https://www.youtube.com/watch?v=ZBwh0Vm4fh4> and in particular the issues raised from minute 42:49 or so.

Good luck
Tzvetan

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20221111/c7a3f586/attachment.htm>