Dear Guillaume,
this fixed the problem. So the hyperparameter estimation was
the reason. Did not think on that. Interestingly I did not encounter
any problems with full brain quantitative images, just with c1
segmentations.
Many thanks,
Niels
-----Ursprüngliche Nachricht-----
Von: Guillaume Flandin [mailto:[log in to unmask]]
Gesendet: Montag, 16. November 2009 13:22
An: Focke, Niels
Cc: [log in to unmask]
Betreff: Re: [SPM] Empty beta image after estimation
Dear Niels,
set option "Variance" to "Equal" (the default is "Unequal") in your
two-sample t-test; this will revert to sphericity assumptions.
Best regards,
Guillaume.
Focke, Niels wrote:
> Dear SPM experts,
>
> I am currently experiencing a quite strange behavior in a 2-group t-test
> ("Basic models", VBM analysis) design with SPM8 comparing a single subject
to
> a group with default settings.
> The design matrix also looks fine after setup (no grey fields in the "beta
> not uniquely specified" row). However after estimation in some cases (about
> 1/3) the first column has an empty beta image (beta_0001) and the "beta not
> uniquely specified" field is now grey (see attached images). Unsurprisingly
> no contrast can be specified and all kinds of error messages are given when
I
> try regardless... I am a bit puzzled why this is happening, again it only
is
> in approx. 1/3 of cases. The other 2/3 of cases work as expected.
>
> Any suggestions are very much welcome.
>
> Niels
>
> ______________________________________________________
>
> Dr. med. Niels Focke
> Dept. of Clin. Neurophysiology
> UMG - University Hospital Göttingen
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|