On 04/27/2017 09:02 AM, stdp82@virgilio.it wrote:
Dear list,
we have several difficulty to understand how I should build fsdg file and we are worried to produce correct results.
On FS site, it is not available an example if we would look for differences among 4 different groups (A, B, C, D; 4 Groups - One Factor/Three Levels - No Covariates). The proposed design with 2 levels and 2 factors is no helpful in our case.
Mathematically, it is the same.
We have 4 groups (1 factor and 4 levels, no covariates) and we our hypothesis (in term of impaiment) is A<B<C<D.
This is not a valid linear contrast since it requires an AND operation (A<B AND B<C AND C<D). You have to use a conjunction analysis in which you create 3 contrasts then conjunct them (with mri_concat --conjunct)
Which should be the group.effect.mtx? How should we look to difference between Group1, Group2, Group3 and Group4? Should we opt for G1+G3 vs G2+G4 or G1+G2 vs G3+G4? Other option?
Please may you argue a rationale?
e.g. Looking your example on three Groups (One Factor/Three Levels), No Covariates, we wonder why if Normals==MCI AND Normals==AD, then it MUST be the case that MCI==AD?
We aim at assess differences by using more complex design (4x2x4), including 4 different structures, looking at right and left structures and having 4 levels (4 different groups).
How should we do, please?
Thank you very much.
Best regards
Stefano
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer