[Eeglablist] Interpolation and ICA Update?

[Matthew Stief]

Greetings!

I was wondering if anyone could weigh in on some (as far as I can tell)
outstanding issues in how to optimally deal with removed channels before
ICA.

The remaining question seems to me to be this: is it superior to A)
interpolate the removed channels, run PCA to appropriately reduce the
dimensionality of the data, and then run ICA, or B) simply leave the
electrodes out and interpolate after running ICA.

With the fact in mind that both the interpolation and the PCA introduce
non-linearities into the data, is it the case that the addition of the new
"clean" interpolated channels provides additional data deriving from neural
sources for the ICA to work with, or does the fact that it's just a
combination of data from other electrodes negate this possibility?  The
negative to the interpolation/PCA approach is clear, and may not be very
substantial, but is there evidence for any positive at all?

-- 
_________________________________________________________________
Matthew Stief
Human Development | Sex & Gender Lab | Cornell University
http://www.human.cornell.edu/HD/sexgender


Heterosexuality isn't normal, it's just common.
-Dorothy Parker
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20120217/2b557f37/attachment.html>