I perform a preproccessing on a task-related EEG data from here in order to reconstruct sources with eLORETA and then create functional connectivity matrices with some FC metrics (wPLI, imCoh and PLV). The dataset’s experiment includes stimuli presentation 2 seconds apart, so I create epochs in [-0.5; 2] second interval around the corresponding event. What I’m concerned about is that, according to this video, I might have edge effects due to dividing the data. On the other hand, I don’t see such artifacts in the time-frequency plane:
All in all, my questions are:
- Should I really care about these edge effects, provided they don’t seem to take place and I use phase-based FC metrics?
- If I should, how could I deal with them? The video suggests creating buffering zones on the edges of an epoch; I guess, it is exactly what the overlap parameter in the make_fixed_length_epochs function is used for. Obviously, there is no such parameter in the Epochs class constructor. Should I use make_fixed_length_epochs instead, relying on the fact that stimuli events are evenly spaced? Is there another way of dealing with such artifacts?