Hi Philip, by using the -m option in design_ttest2, you include individual group mean as the first two contrasts (tstat1 and tstat2) which are not really interesting... What you're presumably interested in are in contrast 3 and 4 (contrasting your two groups), so anything with *tstat3/4.nii.gz. Cheers, Gwenaelle --- En date de : Ven 13.3.09, Philip Koester <[log in to unmask]> a écrit : > De: Philip Koester <[log in to unmask]> > Objet: [FSL] Re : [FSL] Strange t-stats in vbm analysis > À: [log in to unmask] > Date: Vendredi 13 Mars 2009, 16h10 > Hi Gwenaëlle, > > thanks for your quick reply! I ran fslvbm with the standard > scripts from the > homepage (fslvbm_1_bet -b, fslvbm_2_template -a, > fslvbm_3_proc and randomise > -i GM_mod_merg_s2 -m GM_mask -o fslvbm -d design.mat -t > design.con -T -n > 5000 -V). My design.mat and design.con were created with > design_ttest2 > design 20 20 -m, which represent the two groups i want to > compare. I also > tried the non-linear transformation in fslvbm_2_template > which did not > change the results. If i should upload files that could be > more helpful, let > me know. > > thanks a lot, > > Philip