Discrepancy between ICA plots?

✉️ Mailing List Archive (read-only)
1

Dear MNE experts,

I'm currently working on ICA and I'm really confused about an issue.
What makes me uncertain is a (possible) discrepancy between the topomap and
the "epochs image and ERP/ERF" plots of *ica.plot_properties(). *

I'll give you an example: the two figures in this link
<https://drive.google.com/open?id=0B4J_Mny93UZYZXFZRnVjV2w4OG8> show the
ICA component IC#000, the first obtained with
*ica.plot_properties(epochs_clean,
picks=0)*, the second from *ica.plot_sources(epochs_clean)*.
The component in the sources plot shows positive spikes (eyeblinks?) and
these can be noticed also in the top-right plot of the plot_properties.
They are shown in red (positive) -- everything okay thus far.

I'm getting confused looking at the topomap, though. Shouldn't it be
red-ish? It is widespread and blue and it seems not to reflect the data
shown on the right!

The raw data looks ok, I've looked into the documentation and asked an EEG
expert (who's used EEGlab in the past), but we are clueless. I've also
tried a different ICA method, producing the same output. Could the two
plots have different amplitude scales?

Thank you all,

Lorenzo Alfine

MSc student, University of Padova
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20170202/38630fe5/attachment.html

2

hi,

I'm getting confused looking at the topomap, though. Shouldn't it be
red-ish? It is widespread and blue and it seems not to reflect the data
shown on the right!

everything looks ok to me. You see the red blobs over the eyes so it should
match your intuition (more positive over the eye)

Alex

3

Hi Alexandre,
thank you for your reply. What we don't get is why the topomap is mainly
blue. Since the positive peaks are way higher in amplitude than the
negative ones, shouldn't the map be strongly red in the frontal part and
light-blue/white-ish (instead of dark blue) at the rest of the scalp? Am I
overlooking something?
Thank you,

Lorenzo Alfine

4

there is no constraint in ICA that the components are zero mean or
have a specific offset

look at the code if you have a doubt. Bugs are always an option but I
am not convinced there is one.

Alex