Hi Hiroki,

Use the paired t-test FEAT example as the starting point:
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#Single-Group_Paired_Difference_.28Paired_T-Test.29

Then make the following changes:

- In the first EV, instead of 1 and -1, use 0 for the first timepoint, and the actual value of the treatment intensity for the second timepoint (use a consistent scale for all subjects).

- For the exchangeability blocks (.grp file), use one block per subject, each containing the two observations for each subject.

That's it.

All the best,

Anderson

On 20 May 2015 at 03:17, Hiroki Kurashige <[log in to unmask]> wrote:

Hello all,

I'm trying to test the effect of some medical treatment using randomise.

The experiment was paired design, so I obtained fMRI data at pre-treatment and post-treatment.

But, the intensities of treatment were varied between subjects.

Therefore, I try to test using both of  model of  "pre/post categorical variable + treatement intensity variable (0 in the case of pre) + subject categorical variables" and model of "treatement intensity variable + subject categorical variables".

In these cases, how should I make the .grp files?

Thanks,

Hiroki