[Eeglablist] data points and concatenating continuous segments

[Greg Hooper]

I have a continuous data set of 32 channels. Within the set are three
conditions - but these are not ERP type time-locked conditions, but
represent seperate cognitive tasks randomly assigned to be performed
over 4,6, and 8 sec intervals whilst a continuous recording was made.
Should I concatenate the various segments of the three conditions into
their own data sets before performing ICA on each concatenated set
individually? Alternatively should I perform ICA on single 4, 6, and 8
sec segments then average the result obtained across the entire
recording. I worry that disconituity at the boundaries between the
various segments will interfere with the decomposition. On the other
hand concatenation gives the ICA algorithm much more data

thanks, Greg