[Eeglablist] Makoto's preprocessing pipeline - Perform ICA after merged data sets, or perform ICA for individual sets and then merge

[Matsuno, Kevin]

Hi EEGLAB,

I have multiple datasets of the same subject performing the same experiment
and trials.

I would like to append all the datasets and then use machine learning to
classify the signals via BCILAB.

I was wondering in your opinion what is better practice?  To append the
datasets together and then perform ICA.  Or, to perform ICA individually
for each set and then merge?

Do the ICA weights get averaged between the different datasets if I went
the second route?

Or, is there a better way to do this?

Thanks,
Kevin Matsuno