Hi again,
no, R-L flip the 4D image would only lead to the same result, only on the other half part of the volume.
Use the Glm gui, switch from timeseries design to higher-level analysis, set the correct the number of subjects and then use the wizard with single group average: that should create the design.mat and design.con corresponding to the -1 option of randomise when you tested for L>R.
Then, you simply inverse the contrast in the .con to test for R>L.
Cheers,
Gwenaelle
--- En date de : Mar 2.6.09, Siew-Min Gan <[log in to unmask]> a écrit :
> De: Siew-Min Gan <[log in to unmask]>
> Objet: Re: [FSL] Re : [FSL] randomise command after tbss_sym and sample size
> À: [log in to unmask]
> Date: Mardi 2 Juin 2009, 20h04
> thanks very much it works. silly of
> me.
> on the website, it says : To test for R>L, either invert
> the data and
> re-run randomise, or create an appropriate design matrix
> and contrast
> file.
>
> Does inverting the data means fslswapdim input -x y z
> output? Which datas
> should I invert? alternatively, can you kindly give an
> example how to
> create an appropriate design matrix and contrast file.
>
> thanks
>
> Siewmin
>
> Hi Siew-Min,
> >
> >
> >> I then type the command
> >> randomise -i
> all_FA_skeletonised_left_minus_right.nii.gz
> >> -o
> >> mean_FA_symmetrised_skeleton_mask.nii.gz -1 -v 5
> -T
> >
> > your command is not exact: -o is the output of your
> stasticial test not
> > the mask for this test (your mask should be placed
> after -m) and you need
> > to specify the number of permutations (that's why
> nothing ran).
> >
> > Try with this instead:
> >
> > randomise -i
> all_FA_skeletonised_left_minus_right.nii.gz -m
> > mean_FA_symmetrised_skeleton_mask.nii.gz -o
> myresult_L-R -1 -v 5 -T -n
> > 1000
> >
> > Cheers,
> > Gwenaelle
> >
> >
> >
> >
> >
>
|