[Eeglablist] Question about average reference

[Thomas Ferree]

Alan,

>From the perspective of the paper you mentioned, in which the average
reference is sought to approximate the voltage relative to infinity, 16
electrode is too few.  Even if they were distributed uniformly over the
scalp, there is a problem with under-sampling.  If your electrodes are
limited to the back of the head, that sampling is also biased.  The benefit
of using spherical splines is to compensate for the bias by having no
inferior electrodes, but there is a limit to what they can do.

You need not compute the average reference to do legitimate EEG research.
 (In fact, some say that you must not compute the average reference to do
legitimate work, although that is not my personal view.)  Practically
speaking, its main utility is to permit a more direct interpretation of
scalp topographies, and render the weighting of electrodes more similar.
 Depending upon your goals, you may not need this.  Lots of valid research
simply leaves the mastoid reference and states that in the methods.

There are at least two other possibilities if you do want to eliminate the
reference effect.

The surface Laplacian can be computed at all electrodes except those at the
edge of your montage.  That quantity is also independent of the mastoid
reference.  See Nunez P (1981) Electric Fields in the Brain, for
explanation.

Rather than trying to eliminate the reference effect per se, it can be used
with clear intention.  By taking the difference of neighboring electrode
pairs, i.e., forming a so-called bipolar montage, each pair of electrodes
has sensitivity focused in the region under and between that pair.  Bipolar
differences are independent of the original mastoid reference, which cancels
in the subtraction.

Note: If you acquired data with a physically-linked mastoid or ear
reference, none of these techniques works as intended, not to say you can't
try.

Hope that helps,
Tom.

On Fri, Mar 4, 2011 at 3:32 PM, Alan Yi <jackm_ustc at yahoo.com.cn> wrote:

> Hi experts,
>
> I am running a pilot experiment in which I only use 16 electrodes of a
> 64-channel cap. There is not any reference electrode on the cap, so I have
> to re-reference in EEGLAB. Should I just re-reference to the average of the
> 16 electrodes? Will this generate any problems on the result? After all, the
> 16 electrodes are mainly at the posterior of the head. And the paper that
> previously is suggested, Spherical Splines and Average Referencing in
> ScalpElectroencephalography, mentioned that 19 electrodes are too few. So I
> guess 16 should be even worse...But if I don't re-reference to the average,
> would that leads to other problem? Thanks!
>
>
> Alan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110306/59ab0ca7/attachment.html>