Thilo,
Sorry for the delay...
> I have PET-data of 9 subjects (one scan per subject) and want to
> correlate these Scans with a behavioral measure within the framework of
> snpm. Since the function snpm_MSC.m (Multi-subject simple correlation)
> did not work, I tried the same thing for single subjects (the function
> snpm_SSC.m).
I'd interested in hearing exactly how the 'Multi-subject simple
correlation' plugin failed. Didn't save the error messages perchance?
> There I entered "9" as size of exchangability block (option either 3
> or 9). Now I am wondering if this is the right snpm-function (or at
> least a correct work-around) for analyzing the PET data, since the
> function I used is not meant to be used for multi-subject designs.
Sorry for the confusing array of PlugIns. The date from a time when
SnPM was more PET-oriented, and before the summary statistic (Holmes &
Friston) approach to random effects took hold.
Yes, the correct tool for the job is "single subject" correlation model,
where the exchangeability block is all subjects (9). We'll release a
better-named PlugIn soon.
As a preview: We're getting close to releasing SnPM2 version which will
make the post-processing run much faster, specifically cluster size
analysis in particular. The new version writes out images (like SPM2)
instead of huge .mat files and in particular eliminates the gigantic
ST.mat file associated with cluster size inference. (The 'cost' for
these changes is that you'll have to specify your cluster-defining
threshold before you do 'Compute'.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
