[Eeglablist] ERSPs and ICA for movement related potentials

[Marco Mcsweeney]

Dear list

I am trying to run a time-frequency analysis on data collected during a
S1-S2 CNV paradigm with 3 different conditions and 3 different age groups.
I am mainly interested in group x condition differences in mu alpha event
related desynchronisation (ERD) during the anticipatory period prior to S2
onset.

My plan was to 1) run ICA on clean epoched data for each participant
individually to derive components (epochs 5.5 seconds long including a 500
millisecond baseline, 2) use dipfit plugin to locate equivalent current
dipoles with residual variances < 15%, 3) use STUDY to derive clusters
across participants 4) use STUDY to run ERSPs on surviving clusters (based
on scalp map, dipole location, power spectrum and ERSP).

My issue/question relates to the high-pass filter applied to the continuous
EEG data prior to cleaning and epoching the data. The original high-pass
filter was set to 0.05hz and my understanding of ICA is that to get decent
decomposition I should be using a 1hz high-pass filter. I have tried both
and 1hz is definitely better but when using 1hz this completely removes the
CNV ERP.

I understand that I could apply ICA weights trained on a 1hz high-pass data
set to the 0.05hz data set but....

The question is whether using ICA weights trained on 1hz filtered data will
actually pick up components reflecting the CNV which
encompasses frequencies that are lower than 1hz? Also if this is applied to
epoched 0.05hz filtered data does this mean that neural activity relating
to these lower frequencies is shared across different ICs if I use ICA
weights trained on 1hz high-pass filter? i.e. the IC weight matrix derived
from a 1hz filtered data set surely will not be able to clearly identify
independent components present in a 0.05hz high-pass filtered data set?

Also, if I am only interested in the time-frequency domain (mu alpha -
7-12hz band) does this even matter, in which case using a 1hz high pass
filter should be OK.

I have searched online, this mailing list and the literature but am still
not sure about the best way to proceed. Any help or advice on this would be
really really very much appreciated!

Thank you so much in advance.

Best wishes

Marco