Henrik Walter wrote:
> Thanks for your answer. But what I really wanted to know is if it is better
> (more correct) to take the con*images from one Fx-Effects analysis (24
> subjects, both groups) or from two seperate Fx-Effects analysis (12 from one
> group, 12 from the other).
>
> Have you any comments?
>
> H.W.
>
> Russ Poldrack schrieb:
>
> > to do the between-groups analysis, use the two-sample t-test option under
> > basic models and then provide the subject-specific con*.img files for each
> > of your two groups. this will test for differences in that effect between
> > the groups (in either direction depending upon the contrast you specify).
> >
> > cheers
> > russ
>
> --
> Dr. Dr. Henrik Walter
> Abteilung Psychiatrie III
> Universitätsklinikum Ulm
> Leimgrubenweg 12
> 89075 Ulm
>
> fon: 0731/502 1489
> fax: 0731/502 1549
> e-mail: [log in to unmask]
Dear Henrik
As far as I understand the Con images should be the same no matter how many
subjects is in the design matrix. The difference between the two setups lies
within the residual mean square (ResMS) and thus the Ressels per voxel (RPV)
image and the T maps. Because one should only put contrast images into the
second level analysis it makes no difference which setup you use. If you also
want to make fixed effect analysis, you will need the design matrix including
all subjects, and in that case I would just use the big one. If you do not care
about first level analysis and have many subects I would go for one
designmatrix per subject (parameter estimation works faster with smaller
designmatrices).
Torben Ellegaard Lund
M.Sc. Phys. Eng.
Danish Research Center of Magnetic Resonance
Hvidovre Hospital
Kettegaard Alle 30
2650 Hvidovre
Denmark
Phone: +45 3632 2978
Fax: +45 3647 0302
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|