Hello!
I wanted to ask if there are any special pre -processing steps for analyzing coherence for small frequencies for a 64-channel EEG recording of a cognitive task?
I would recommend avoiding most pre-processing steps documented in the tutorials for ERP analysis:
- Do not apply any frequency filter - band-pass or notch: just ignore the unwanted peaks in the coherence spectrum
- Do not run any ICA, SSP, re-referencing or other spatial filters, as these methods introduce a lot of correlation between the signals
- Mark as bad all the artifacts with a slow temporal dynamic (< 8Hz, e.g. blinks). Noise at frequencies higher than the frequency of interest should not be a problem.
- A peak between 1 and 2 Hz might be related with the heartbeats.
@Sylvain @Marc.Lalancette @Raymundo.Cassani
Any other recommendation?
Sensor space or source space?
Sensor space.