Hi,

I would suggest calculating the paired differences for each subject in separate lower level models (one for each subject) and then running a 2-sample t-test using these differences (now 1 entry per subject).  Unless you have variance estimates for each measure,  you can just use fslmaths to calculate the paired differences for each subject.  So if you have 40 subjects you'll calculate 40 differences and then enter these into a 2 sample t test.

As far as I know it is not possible to get randomise to properly permute the data otherwise, since you'd need it to permute the pairs of data for each subject together.  This also greatly simplifies your group model.

Cheers,
Jeanette

On Tue, Jan 17, 2012 at 2:39 PM, Leslie Engineering <[log in to unmask]> wrote:

Or is there a website I could be referred to? I tried the three designs that made sense to me, but fsl doesn't seem to like them (as per the warning in my first email).  I thought a column where controls were 1s, a column were patients were 1s, a column where visit 1 were 1s, and a column where visit 2 were 1s. 

Maybe it was my contrasts?

cv1: control visit 1

cv2: control visit 2

sv1: patient visit 1

sv2: patient visit 2

My contrasts were:

                  c       s       v1       v2

cv1>cv2      1       0        1         -1

sv1>cv1      -1       1        1         0

.

.

.

ect. 

On Tue, Jan 17, 2012 at 12:02 PM, Louie Bird <[log in to unmask]> wrote:

Thanks. How would you make a design contrast for 2 groups with two visits each?

Thanks so much

On Jan 17, 2012, at 11:37 AM, Matthew Webster <[log in to unmask]> wrote:

> Hello,
>          This message is generally a sign that there is an issue with the design/contrast - it's probably best to kill the randomise task here and recheck the design. Alternatively you could try re-running the contrast with a different clustering option than TFCE, but I wouldn't expect the output to have any significant voxels.
>
> Many Regards
>
> Narrgew
>
>> After successfully running TBSS I wanted to investigate group comparisons. I made a design matrix/contrast by using glm_gui. I implemented those files in randomise. The first time I ran it everything seemed to work fine but I realized that my design was flawed. I made a new design and attempted again, but received the following message:
>>
>> randomise options: -i all_FA_skeletonised.nii.gz -o ../../TEST/tbss -m mean_FA_skeleton_mask -d ../../TEST/Test.mat -t ../../TEST/Test.con -n 500 --T2 -V
>> Loading Data:
>> Data loaded
>> 7.79255e+14 permutations required for exhaustive test of t-test 1
>> Doing 500 random permutations
>> Starting permutation 1 (Unpermuted data)
>> Starting permutation 2
>> Warning: tfce has detected a large number of integral steps. This operation may require a great deal of time to complete.
>>
>>
>> Nothing has changed in 3 days. Can someone tell me what might be wrong?
>>
>> Thanks
>> Lou