Thank you. I have required randomise to generate t-maps as well, but there is one thing that still confuses me, and it might be because I am doing something wrong... How can it be that when looking through fslstats -R (to see what the maximum t is in the map), I get new "significant" results (over a threshold of 0.95) in the t-maps, i.e. a cluster of 40 when looking at one t-map whereas the original f-map only had a cluster of 6? Shouldn't it be the same amount of voxels?
As it could be because I haven't created correctly my contrasts:
design.mat (it's a simplified example): EV1 is the group column and EV2 is the covariate
EV1.... EV2
-1.... 85
-1.... 71
-1.... 65
-1.... 98
-1.... 82
1.... 95
1.... 75
1.... 83
1.... 82
1.... 70
design.con
1 0
-1 0
And I ran randomise using the -D -T -n 10000 options
Can it be that the fslstats -R option wouldn't give the same results since we're talking about f-maps in one case and t-maps in the other? I do realize that if 1 cluster or 3 clusters are identified in the f-maps, there will be 1 or 3 clusters also identified in the fslstats command ran on t-maps. If so, considering the matrix, what would be the direction of a [1 0] contrast? That the 1st group presents with negative values, i.e. locus inside the boundary, i.e. surface displacement towards the inside, i.e. "atrophy"?
As you can see, I am... puzzled!
Thank you so much
Shady
|