Dear Laura,
...
> The question is about orthogonality of contrasts when looking a group
> by task interaction effects in a second level analysis.
>
> Suppose that you have a simple ANOVA with two columns one for a
> control group and one for a patients group
>
> Contrast images here represent signal changes of a task minus baseline
> comparison (say reading miinus fixation)
>
> My recollection is that the two contrasts
>
> 1 -1 the group x task interaction effects
>
> and
>
> 1 1 the main effect of activation across the two groups
>
>
> are orthogonal
These are orthogonal contrast weights, but not necessarily orthogonal
contrasts. However, if the first-level regressor is identical in all
subjects, and you have the same number of subjects in the control and
patients groups, then the contrasts are orthogonal.
> and therefore the nominal significance in the first SPM
> ( the 1 -1 ) should be the square
> for the voxels that are also significant for the 1 1 contrast
> (voxel identified by inclusive masking)
I must confess I do not understand what you mean by square of nominal
significance.
Best wishes,
Roberto Viviani
Dept. of Psychiatry,
Univ. of Ulm,
Germany
|