Hi Thomas,
...or just use "Multi-subject: cond x subj interaction 7 covariates", and
put all the subjects in together. This will ask you less questions, but
effect the same result as Karl's suggestion.
Although you have two groups, you only want the individual subject level
contrasts, which you then will assess at the second level. So, the group
membership isn't important in the first level of the analysis since the
model fits each subject separately.
As Karl notes, all effects must be fitted as interactions with the subject
effect to ensure subject separability.
-andrew
At 12:22 12/07/2000 +0100, Karl Friston wrote:
| At 13:13 12/07/2000 +0200, Thomas Stephan wrote:
| > when we do a 'multi-group conditions and covariates' analysis, we can
| > not produce the subject specific contrast images, because the columns
| > of the design matrix contain the different conditions for all
| > subjects. How can we obtain the contrast images to enter into the
| > second level analysis ? Do we have to use the results from the single
| > subject analysis ?
|
| You could proceed by treating each subject as a separate group.
| If you are using covariates ensure you select 'covariate x subject
| interactions'.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|