Dear Philippe,
someone has suggested to me to use still another approach, which I
presonally incline to prefer. Maybe somebody else could comment on this.
1) Compute for each subject the two contrasts A and B separately
2) Compute the second level RFX stats for the two contrasts separately
3) Write the thresholded statistical maps obtained by step 2 for
contrast B and call it say MASK.img
4) Small volume correct (SVC) the RFX stats of contrast A obtained by step
2 with the MASK.img generated in step 3.
However, you have to make sure that contrasts A and B are fully
orthogonal, because SPM99 won't alert you in case they are not and produce
invalid statistics.
Best regards,
Marco
Marco Tettamanti
Istituto Scientifico San Raffaele
Reparto Medicina Nucleare
Via Olgettina 60
I-20132 Milano
Italy
Tel. 0039 - 02 - 26 42 34 60
Fax: 0039 - 02 - 21 71 75 58
email: [log in to unmask]
|