Hello,
Thank you so much.
I'll try this substraction method!
Best regards,
christophe
Selon Jeanette Mumford <[log in to unmask]>:
> Hi,
>
>
> >
> > I am unfortunately lost.
> >
> > I would like to compare two group-ICA components ( a same group but two
> > different tasks) by randomization.
> >
> > I have two questions:
> >
> > 1. What must contain the 4D volume to be used? Must it be built with
> > dr_stage2_ic.nii.gz of the concerned components? and How?
> >
> > The best option is to identify the pairs of ICA components and subtract
> them using fslmaths, within subject, to create a set of differenced images
> (1 per subject). Then you would enter these into randomise as a one-sample
> t-test (1-sample t-tests of paired differences is a paired t-test).
>
> Somebody else can correct me if I'm wrong, but from reading
> this<http://www.fmrib.ox.ac.uk/analysis/dualreg/> it
> would appear that you'll have a dr_stage2 file for each subject and,
> depending on how many components you chose to run the dual regression on,
> this may be a 3D or 4D image. In the 4D case you will need to split with
> fslsplit and extract the spatial map associated with the component of
> interest. Calculate your paired differences, etc. To get the 4D file for
> randomise you will use fslmerge -t to concatenate the subject-specific
> paired difference images.
>
>
> > 2. Are the design files built with FEAT: high level statistics two-groups
> > paired?
> >
>
> I'm not sure what you're referring to, but it is either a 2-sample t-test
> or the version of a paired t-test that will not work in randomise.
>
> Hope that helps,
> Jeanette
>
|