Dear fsl experts,
Basic randomise output of a TBSS analysis are maps with p-values. If you want to output effect-sizes, in this case the betas, or "Parameter Estimates", you need to calculate them yourself. This was pointed out by Prof. Smith (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0701&L=FSL&P=R26572&1=FSL&9=A&I=-3&J=on&X=1C67F64AE3865282CC&Y=v.koppelmans%40erasmusmc.nl&d=No+Match%3BMatch%3BMatches&z=4).
My model has one contrast:
Group A < Group B
It has several covariates.
After calculating the PE (film_gls -v -noest all_FA_skeletonised.nii.gz design.mat), I masked the PE with a mask based on the TFCE corrected p-value-maps (fslmaths tbss_ctfce_corrp_tstat1 -thr 0.949 mask_1).
The thing that I do not understand is the following:
within the masked PE file (fslmaths PE1 -mas mask_1) there are both voxels with negative AND positive values.
As I set my contrast to detect voxels with lower FA in Group A then Group B, I expected only negative beta values. At least within the mask.
>> Is there something wrong with my calculations or with my interpretation? Can somebody explain why there are still positive voxel (beta) values within the mask?
Kind regards,
Vincent
|