On Wed, 5 Sep 2007 08:34:05 +0200, Volkmar Glauche
<[log in to unmask]> wrote:
>You should read (and perhaps also use) spm_config_contrasts to create your
>contrasts - it should be fairly easy to create the appropriate job
>structure in MATLAB and then call spm_jobman('run', ...) to compute your
>contrasts.
I've been trying to figure out how to make contrasts
in SPM5 in batch mode, without any GUI-clicking.
For preprocessing, specifying and estimating models,
the SPM5 save-a-mat / load-the-mat-then-insert-new-info approach
works well, and is much nicer than things used to be in SPM2.
After having done one subject by hand
and saving their design in a .mat file,
that .mat file contains exactly the right job structure.
However, it's not clear to me how to make the right job
structure for batching contrasts. From the previous discussions
of this on the list, it looks as though people have gotten this
to work, using some modification of spm_config_contrasts.
I couldn't find any actual posted code, though.
If anyone has an example of such job-creating-and-running code,
then I'd be very grateful if they could post it.
I ended up hacking some code to make contrast-images
by directly making weighted sums of the beta images.
That's ok for getting the contrast-images for random-effects,
but I don't end up with spmT images for each subject,
or per-subject contrasts which SPM Results can recognise.
Many thanks,
Raj
P.S. Slightly off topic, but if anyone could give a jargon-free
description of the current status in SPM of whether you
can do an FSL-esque multi-subject mixed model
which uses each subject's variance info as well as their
contrast image, then I'd be very grateful.
Emphasis on jargon-free :-)
|