Does anyone know of any sensible formal criterion for identifying "bad" EEG channels? I am wondering whether some min-max (in either time or frequency domain) over a specified period compared to the average/standard deviation across all other channels might be appropriate?