Hi,
the full tensor approach assumes that designs are consistent between
runs, so across the temporal domain and the session/subject domain a
single design should do the trick.
When using the concatenation approach I've chosen to perform a simple
rank-1 approximation at the end, i.e. show the single largest Eigen-
time course in the report (as opposed to plotting all time courses or
plotting the single long time course), mainly for reasons of
simplicity (plotting the long concatenated time course seems
uninformative in most cases), so the 'right' design matrix would be
the one containing the Eigen-EVs of al design matrices I'm afraid.
Despite what's in the report you do have access to all time courses,
they are contained in melodic_mix and are also contained in the
t*.txt files in the report directory. We also now supply a simple
command line tool fsl_glm which can perform simple OLS multiple
regression of text matrices (simple ascii or output from Glm) against
text matrices, 4D data against text matrices and 4D against 4D
(spatial regression). You can use this to regress the separate design
matirces against estimated time courses and get hold of GLM betas,
Ts, Zs etc that way.
hope this helps
Christian
On 16 Aug 2007, at 20:31, H.Steven Scholte wrote:
> Dear FSL creators,
> I was wondering how to enter a design for the multi-session ICA
> (and the
> other ICA option).
>
> For a single subject you just enter the contrast and design files
> from GLM.
> What to do in the multi-subject case when the design files are
> different (in
> that case I should use the temporal concatenation?), also
> concatenate the
> model files?
>
> yours sincerely, Steven
> ps
>
> many thanks for making such a great program
____
Christian F. Beckmann
University Research Lecturer
Oxford University Centre for Functional MRI of the Brain (FMRIB)
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
[log in to unmask] http://www.fmrib.ox.ac.uk/~beckmann
tel: +44 1865 222551 fax: +44 1865 222717
|