Dear FSL experts,
I am using two-sample paired t-test using Randomise using TFCE (n=15 conditions=pre, peri). When I threshold at 0.95 and 1, there are no significant voxels/clusters at all. This is against what I expected, as normal higher-level FEAT (with clustering threshold) returns several significant voxels/clusters - It seems strange to conclude that none of these clusters are significant because those were not deemed significant in Randomise.
Even in one sample t-test for each condition, for the regions deemed significant in higher-level FEAT contrast images, there is a huge difference in terms of significance, further supporting that the regions should be deemed significant in contrast images.
Can anyone provide me some insight?
Thanks,
Here are the commands I used. (For design.mat, the entries followed the conventional design matrix scheme for paired t-test)
randomise -i LC_peri_vs_pre.nii.gz -o LC_peri_vs_pre -d design.mat -t design.con -n 5000 -T
#design.grp
/NumWaves 1
/NumPoints 30
/Matrix
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
#design.con
/ContrastName1 condition A > B
/ContrastName2 condition B > A
/NumWaves 16
/NumContrasts 2
/PPheights 2.000000e+00 2.000000e+00
/RequiredEffect 2.368 2.368
/Matrix
1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
-1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
|