Hi - the main thing here would appear to be that a very low (or zero)
number of valid voxels is being fed into flame - I suggest that you look
at the higher-level feat web page registration summary page to check
registrations and mask overlaps feeding into this.
Wrt the 2-group variances setups, you need "separable designs" - see the
bubble popup when you hold the mouse over the "number of groups" button.
Hope this helps - Steve.
On Wed, 3 Sep 2003, Goekoop, R. wrote:
> Dear FSL-users,
>
> In a design where 30 subjects were scanned at baseline and 3 months after
> placebo (N=15) or medication (N=15) intake (a 'nested' design), I performed
> higher-level analysis in FEAT to extract the effect of medication vs.
> placebo. For this, I first analysed the entire population without specifying
> groups, comparing the effects before and after treatment using 2 EVs coding
> for contrasts (session 2 > session 1). One EV explained signal variance
> (before and after treatment) for the placebo group, the other EV examined
> this same effect for the group receiving medication (a contrast between the
> two EVs would yield the effect of medication vs. placebo, corrected for
> baseline activation levels). No problems occured here.
>
> Afterwards, I felt it might be more accurate to model separate variances for
> both groups by specifying group memberships. This analysis ran for about an
> hour, after which the following error occurred:
>
> 1DOF cannot be zero or negative!
> ...etc.
>
> stdtr domain error
>
> ndtri domain error
>
> ...etc (see attachment for an excerpt of the report.log file).
>
> A previous mail in the archives reported a similar problem when groups were
> separated (Zrinka Bilusic; Sat, 29 Mar 2003 06:53:27). If I read it well, it
> seems that if registration has not been carried out correctly in some cases
> (i.e. the entire brain has not been scanned as judged from the individual
> brain masks) the splitting up a population into two groups may create two
> populations that differ in the number of voxels analysed for a particular
> location. Am I correct to think that this asymmetry could cause the problems
> reported here when these data are combined in the final stages of a
> multi-group higher-level analysis?
>
> Thanks,
>
> Rutger Goekoop
>
> Drs. R. Goekoop, MD.
> Department of Neurology
> Vrije Universiteit Medical Centre
> P.O. Box 7057, 1007 MB
> Amsterdam, the Netherlands
> Phone: +31 20 444 0316
> E-mail: <mailto:[log in to unmask]> [log in to unmask]
>
>
>
Stephen M. Smith MA DPhil CEng MIEE
Associate Director, FMRIB and Analysis Research Coordinator
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|