Dear Paul,
This might have to do with too disparate head locations between your
subjects. Group inversion can only work if there is enough common
information picked by the sensors from all subjects. When the head
locations are not consistent then as you add subjects the common
subspace reduces in dimensionality until there is nothing left. Here
at the FIL group inversion has been done for 18 subjects and probably
there is no hard limit if you control the positioning well.
Best,
Vladimir
On Mon, Aug 8, 2011 at 5:00 AM, Paul Sowman <[log in to unmask]> wrote:
> I have a problem doing group inversion on 12 subjects. All have 6 conditions that are 1000ms in duration. When I try to do the group inversion with 5 or more of the subjects included I get 0 spatial and 0 temporal modes in the feedback and log evidence & hyperparameters:
> NaN NaN
> NaN NaN
>
> If I do the same group inversion with 4 or less subjects included everything seems to work fine. Any ideas??
>
> Thanks, Paul
>
|