Dear Oliver,
>
> I tried to do a conjunction analysis at the second level as proposed in
> the email from 09/02/99. I took the estimates of the simple main
> effects to the second level, used 'multiple regression', but,
> unfortunately, the contrast manager did not allow me to enter contrasts
> testing for the significance of these main effects in the group, i.e.
> the effects of every single regressor (for example [1 0 0]). Only
> 'balanced' contrasts (for instance [1 -1 0]) were accepted. Did I make
> something wrong? Do you think that there is another way to perform
> conjunctions (AND masking*) at the second level, for example by use of
> condition-specific images as regressors? (*I would like to mask the
> interaction contrasts of two simple main effects [-1 1 1 -1] with the
> main effects itself [-1 1 0 0].)
I suspect that you have modeled a constant term at the second level (which
should be omitted).
With best wishes - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|