Dear Vasilis,
Have a look at this document and linked articles for advice on how to
specify design matrices for mixed designs:
https://en.wikibooks.org/wiki/SPM/Group_Analysis
You can use a mask obtained from one contrast to restrict the search
space for another contrast provided that the spaces the contrasts span
are orthogonal.
Best regards,
Guillaume.
On 15/11/2019 16:44, Vasilis Ioak wrote:
> Hi,
>
> I'm quite new to SPM and fMRI analysis so the answer to my problem might be quite obvious to most users.
>
> I have data from an n-back paradigm in patients (SCZ) vs healthy controls (HC). I am interested in the contrast HC > SCZ for the within subject contrast 2back > baseline. I have created a flexible factorial design matrix with the factors 1)group (2 levels; HC, SCZ) and 2)n-back (3 levels; 1back, 2back, 3back). So to explore the HC > SCZ for 2back > baseline, I use a T contrast [0 0 1 -1 0 0] and get frontoparietal differences in activation (p uncorrected).
>
> I've noticed, however, if I open the F contrast with the average effect of nback [ 1 1 1 1 1 1 ], and then apply the T contrast HC > SCZ, 2back > baseline as a mask, I get pretty much the same frontoparietal areas but with much more significant p values at voxel-FWE correction.
>
> So far, I don't fully understand the nature of masking, so I'm wondering whether it is valid to use my T contrast of interest as a mask on the average effect of n-back (F contrast). Is this considered a ROI analysis as opposed to whole-brain?
>
> Many thanks,
>
> Vasilis
>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|