Dear Mark,
thank you very much for your quick feedback.
In the meantime, I went on reading and studying,
and eventually decided to perform a 1-sample ttest for each group, to get the mean effect whithin the group, with, for example, the command
randomise -i 1sample_seed1_pz.nii.gz -o 1sample_Ttest_seed1_pz -1 -v 2.5 -T
In 1sample_seed1_pz.nii.gz i concatenated the cope images for all patients obtained from level 2 of the analysis,
and I decided to perform variance smoothin considering the small groups.
And then perform the 2-sample ttest to compare the groups with the command
randomise -i 2sample_seed1.nii.gz -o 2sampleTtest_seed1.nii.gz -d design.mat -t design.con -n 5000 -T -v 2.5
In 2sample_seed1.nii.gz i concatenated the cope images for patients followed by the cope images for controls,
and design.con now is the following:
/ContrastName1 group A > group B
/ContrastName2 group B > group A
/NumWaves 2
/NumContrasts 2
/PPheights 1.000000e+00 1.000000e+0
/RequiredEffect 2.801 2.801
/Matrix
1.000000e+00 -1.000000e+00
-1.000000e+00 1.000000e+00
Hope this makes sense.
Thank you all,
Stefania
________________________________________
Da: FSL - FMRIB's Software Library [[log in to unmask]] per conto di Mark Jenkinson [[log in to unmask]]
Inviato: venerd́ 14 marzo 2014 9.50
A: [log in to unmask]
Oggetto: Re: [FSL] randomise resting-state data
Hi,
That generally looks fine, except that you should feed the COPE images into fslmerge and randomise, not the zstats.
Also, you should not use -D in the randomise call, as you are interested in the mean values, so you certainly don't want to demean the data!
All the best,
Mark
On 13 Mar 2014, at 14:14, Stefania Evangelisti <[log in to unmask]> wrote:
> Dear fsl experts,
> I'm quite new here
> and I'm writing to ask you an opinion about the following analysis:
>
> resting state data, 9 patients and 9 healthy controls,
> each subject has got 2 runs.
> Following http://fsl.fmrib.ox.ac.uk/fslcourse/lectures/practicals/feat2/index.html#multisession I ran level 1 (seed-based functional connectivity for each run) and level 2 (mean across each subject's runs) on my data.
> As for level 3, instead of the cluster thresholding that can be set from the Feat GUI, I'd like use randomise on these data to perform a 2-sample unpaired T-test to investigate differences in connectivity between the two groups.
>
> The steps that I'd do are the following:
>
> - concatenate (with fslmerge) in a 4D image (4Dimage_ztats) the zstat images of each subject (obtained in the 2nd level of the analysis)
>
> - use the following design.mat:
> /NumWaves 2
> /NumPoints 18
> /PPheights 1.000000e+00 1.000000e+00
>
> /Matrix
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 1.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
> 0.000000e+00 1.000000e+00
>
> and the following design.con
>
> /ContrastName1 group A > group B
> /ContrastName2 group B > group A
> /ContrastName3 group A mean
> /ContrastName4 group B mean
> /NumWaves 2
> /NumContrasts 4
> /PPheights 1.000000e+00 1.000000e+00 1.000000e+00 1.000000e+00
> /RequiredEffect 2.801 2.801 1.981 1.981
>
> /Matrix
> 1.000000e+00 -1.000000e+00
> -1.000000e+00 1.000000e+00
> 1.000000e+00 0.000000e+00
> 0.000000e+00 1.000000e+00
>
>
> - run the following command:
>
> randomise -i 4Dimage_ztats -o ttest_randomise -d design.mat - t design.con -n 5000 -D -T
>
>
> Am I making any conceptual mistake due to my inexperience?
> Thank you very much for any suggestion.
>
> Stefania
|