[Eeglablist] odd channel properties - issues with the filtering

[Mar Cordero]

Dear EEGLAB community,

I am in the first year of a PhD program and I am analysing the EEG data
from my first experiment. After reading several ways to do the
preprocessing, I have done it as follows:
1- Import data
2- Filter 0.1 – 30 Hz
3- Left the default channel locations
4- Visual inspection of bad channels and bad segments
5- Epoch extraction (from -200 to 1100 ms)
6- Baseline correction for -200 to 0
7- Visual inspection of trials (and rejection of bad ones)
8- Create a version with a high-pass filter of 1 Hz
9- Run ICA with Picard on 1 Hz-filtered version
10- Save weights temporarily and transfer them to 0.1-Hz-version
11- Check IC components by map and also use the Plot>components activations
to check for the voltages of each component

When looking at the channel properties, and also at components after
running ICA, I see slight differences among the activity power spectrum of
ALL the components and subjects. I am not sure whether it has to do with an
issue in the filtering step. I send you attached some screenshots of how
continuous raw data look like, as well as trials after epoch extraction.
There is one picture of how channels are before applying the filter (images
upper row) and after applying it (images in the lower row). And finally the
properties of some of the ICA components.

I do all the steps through the EEGLAB GUI in Matlab. The software we have
in our lab is Neuroscan (with the Curry 8 Data Acquisition package) and we
have two SynAmps 2/RT since we have 64 and 128-electrode caps, although for
my experiment I only used 64 channels.

Do they look odd? Is there another way of filtering I should use?

I would appreciate any comments and recommendations you may have for me. I
have been struggling to search for this kind of issues but it has been hard
to find anything related.

Thank you a lot in advance.

Regards,
Mar