Thank you Dr Anderson, this is very helpfull! ...
Hi Jon,
These results look fine. PALM saves p-values, not 1-p. Consider using the option "-logp" (recommended) or "-save1-p".
All the best,
Anderson
On 23 June 2017 at 18:12, John Anderson <[log in to unmask]> wrote:
Dear Dr Anderson,
I have two groups of subjects (40 patients and 20 HC). In order to study, voxel wise, the difference between the groups in PET signal. I merged the final output of the PET data using "fslmerge". Then I built design matrix and I fed it to PALM as follow:
$PALM/palm -i PET_data.nii.gz -o vw -d design.mat -t design.con -m MNI152_T1_2mm_brain_mask.nii.gz -n 5000 -T -C 3.1 -corrcon -noniiclass
I ran the analysis previously using randomise as follow:
randomise -i all.nii.gz -o vw -d design.mat -t design.con -m MNI152_T1_2mm_brain_mask.nii.gz -n 5000 -T
and I got the *tfce* statistical map of interest which I visualized it using "fslview" (min =0.95 and max=1), So I know exactly where is the difference between the groups and how it look like..
Regarding PLAM, I can't really understand the *tfce* output of PALM or how to visualize it. kindly see attached. I threshold the *tfce* statistical map similarly to what I have done for the output *tfce* images in randomise, but what I see here is that the areas of the significant difference disappeared and the areas of no difference persist. What I am doing wrong?
I highly appreciate any guidance...
Cheers,
Jon
|