[Eeglablist] Pre-clustering with two dipoles

[Francesco Marini]

 Dear list members,

I would like to discuss an aspect of pre-clustering when two dipoles are
present in the data.

The commented code in std_preclust suggests that two dipoles are handled by
selecting the leftmost one:

[std_preclust lines 331-332]
% find the leftmost dipole
[garb ldip] = max(ALLEEG(idat).dipfit.model(icomp).posxyz(:,2));

However, I do not understand the approach because the coordinate that
determines left/right in Tal & MNI systems is stored in the first column
and not in the second:

K>> ALLEEG(idat).dipfit.model(icomp).posxyz

ans =

  -37.8950  -74.0210   10.5650
   37.8950  -74.0210   10.5650

My understanding is that in order to pick the leftmost dipole one should
look at the first column rather than at the second, and because the
leftmost dipole has a negative coordinate, I would look for the min rather
than for the max (again in Tal/MNI, this may not be correct for other
coordinate systems).

So I'd have though of using something like (differences in bold):

% find the leftmost dipole
[garb ldip] = *min*(ALLEEG(idat).dipfit.model(icomp).posxyz(:,*1*));

I realize that choosing either the left or the right is arbitrary and
wouldn't make much difference after all. But I would like to understand why
the max along the 2nd dimension gives the leftmost dipole.

As a suggestion, perhaps choosing the dipole with the largest momentum may
have a better theoretical justification.

Francesco

P.S. The eeglab wiki might need to be updated because it reads "If one
component is modeled using two symmetrical dipoles, pop_preclust. will use
the average location of the two dipoles for clustering purposes (Note: this
choice is not optimum)."


-- 
Francesco Marini, Ph.D.
Postdoctoral Research Associate
UN Reno & UC San Diego
Personal website: fmarini.weebly.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20180412/4719d6a8/attachment.html>