Dear Jesper, Eric and Mike,
Many thanks for your kind assistance on the ‘segmentation fault” which
we first presented to the SPM list on
(http://www.mailbase.ac.uk/lists/spm/2000-02/0074.html).
> >I also tried with a bigger group size (30) and had the “segmentation
> >fault” message too.
> >
> This seems to be a software problem and is unlikely to change with group size.
It took us quite a long time, but we have just found out that by
augmenting the actual stacksize (8192 kbytes) of our SUN workstation
with the UNIX command: limit stacksize 65536, we have managed to
solve the problem in SPM99. I suppose that the problem didn't appear
with SPM96 because this version demanded less resources.
As it took us a long while to sort that out, we weren't able to test in
SPM99 the correct design you suggested for our particular study with one
scan per subject ("Single-Subject: covariates only"). We have done that
now, and it works fine.
>> I have no problems with “Multisubject: covariates only” in SPM96
>> (Proportional Scaling/ No Grand Mean Scaling/ Gray Matter threshold:
>> 0.8)
>
> I must say this surprises me somewhat. It shouldn't be possible to analyse
> that way with only on scan per subject since, as Eric pointed out, the
> subject effects are completely confounded with your covariate
Many thanks, and sorry for taking your time!
With best regards.
Griselda
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|