On 10 Apr 2018, at 09:43, Wu Yulun_ <[log in to unmask]> wrote:

Dear FSL experts,

I have used FSL for randomise:

randomise -i 4Ddata.nii.gz -o tbss -m mask.nii.gz -d design.mat -t contrast.con -n 500 --T2

The 4Ddata contains 55 images, and the design and contrast txt file is attached. For the covariate, it contains many decimal digits. When I do this in FSL 5.0.9, there is an error in the second contrast: unpermuted image contains no positive value, so a blank images is created. (but if I use the covariate with integral number, no problem) But when I do this in FSL 5.0.11, it's OK. Do you know why is that?

Yours,
 
Yulun Wu

Faculty of Applied Sciences

Delft University of Technology

<designmotionsingle.txt><contrast1.txt>