Hi Alena,
So you have 2 groups, 2 treatments, 1 continuous nuisance and 2
measurements per subject. While this can go straight into randomise with
the definition of exchangeability blocks, this can be somewhat
complicated and error prone. Moreover, this isn't currently described in
the FSL wiki. As of today, the recommendation, for ease of use, is to
subtract the 1st timepoint from the 2nd (i.e., t2-t1, with fslmaths) for
each subject, and use the difference as the input image for randomise.
Then model as a 2x2 ANOVA
(http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-factors_2-levels_.282-way_between-subjects_ANOVA.29),
but adding the extra nuisance regressor (i.e., ANCOVA).
However, it seems your nuisance isn't an ordinary one, like age, sex, or
IQ, but compliance to the treatment. If the hypothesis that exercise
and/or therapy change the response over time is correct, the compliance
may be considered the "strength" of the treatment, something
interesting, rather than a nuisance that you'd want to get rid of. I
think that I'd instead use it as an independent variable of interest,
and assemble the design as in this other example:
http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#Two_Groups_with_continuous_covariate_interaction.
Hope this helps.
All the best,
Anderson
On 15/01/2014 12:10, Alena Chrastek wrote:
> Dear FSL users,
> I have a question about setting up the contrast in GLM-repeated measures design.
> I have two groups patients and controls, both group were randomised into exercise or occupational therapy. Every subject were measured twice in the beginning of the study and after 6 months. I want to include compliance as covariate as well.
>
>
> Run(measurement1or 2) Group(patient/control) randomisation(exercise/occ) compliance
> 1 1 0 50
> 1 0 1 75
> 1 0 0 25
> 1 1 0 100
> 0 1 0 50
> 0 0 1 75
> 0 0 0 25
> 0 1 0 100
>
> I want to study whether there is an difference in time by randomisation and time by group by randomisation in white matter parameters in FSL.
> Could someone help me with setting up the contrast to achieve this?
> Many thanks for answer
> Alena
|