In my analysis of resting state EEG data I would like to have relative long chunks of data to compute PSD and some metrics from it; I am using autoreject to find and reject moment with big artifacts before ICA (the data is really noisy) and and I would like to retain more epochs for later analysis.
my question is if is ok to use overlapping data (1,5s in total between 5s epochs) for ICA as implemented in MNE?
If I would use non-overlapping epochs I can apply my fitted ICA on my raw file? What if delete/drop some epochs?