Dear FSL team,
I am doing an analysis using TBSS looking at the effect of a treatment in a group of subjects longitudinally. There is no control group as everyone is on medication.
After a pipeline to run the longitudinal analysis I ended up with an all_baseline_fa_skeletonised.nii.gz and all_finalvisit_fa_skeletonised.nii.gz (and same for MD, AD, RD) in the same space
Since I am interested to see if there is any significant change over time I have subtracted the finalvisit skeletonised image from the baseline visit and obtained an all_fa_substract.nii.gz image.
In the matrix I added 4 EV, one variable for the group mean and then three covariates as below
Group mean age sex site
1 17 1 2
1 23 0 3
1 15 1 1
1 45 1 1
Then I would do only one contrast to look at if there is any significant change at all
EV1 EV2 EV3 EV4
C1 1 0 0 0
and I ran randomise
randomise -i all_fa_substract.nii.gz -o longitudinal_fa -m mean_FA_skeleton_mask -d longitudinal.mat -t longitudinal.con -n 5000 --T2 -V –D. (also tried without -D, results unchanged).
Finally, I multiplied the all_fa_substract.nii.gz by -1 and repeated the analysis.
To double-check that the results were correct I merged the baseline images with the finalvisit images so I would have in a single file first all the baseline images followed by the final visit images
Then I would set up the following matrix (example for four participants)
Group EV1 EV2 EV3 EV4 EV5 age sex site
1 1 1 0 0 0 17 1 4
2 1 0 1 0 0 18 0 3
3 1 0 0 1 0 19 1 2
4 1 0 0 0 1 45 1 1
1 -1 1 0 0 0 17 1 4
2 -1 0 1 0 0 18 0 3
3 -1 0 0 1 0 19 1 2
4 -1 0 0 0 1 45 1 1
And for the contrasts
EV1 EV2 EV3 EV4 EV5 age sex site
C1 1 0 0 0 0 0 0 0
C2 -1 0 0 0 0 0 0 0
And randomise again:
randomise -i all_fa_blANDd421.nii.gz -o longitudinal_fa_two_paired -m mean_FA_skeleton_mask -d longitudinal_two_paired.mat -t longitudinal_two_paired.con -n 5000 --T2 -V –D
I also tried
randomise -i all_fa_blANDd421.nii.gz -o longitudinal_fa_two_paired -m mean_FA_skeleton_mask -d longitudinal_two_paired.mat -t longitudinal_two_paired.con -e longitudinal_two_paired.grp (with all the -n 5000 --T2 -V –D (in the .grp file here I included the numbers as in the first column.
However, to my terrible surprise, the results are completely different!
I have repeated the analyses and checked obsessively but the results do not match!!
Of note, I have also tried to run an unpaired t test as below
Group EV1 EV2 age sex site
1 1 0 17 1 4
1 1 0 18 0 3
1 1 0 19 1 2
1 1 0 45 1 1
1 0 1 17 1 4
1 0 1 18 0 3
1 0 1 19 1 2
1 0 1 45 1 1
And for the contrasts
EV1 EV2 age sex site
C1 1 -1 0 0 0
C2 -1 1 0 0 0
And again, results similar to the paired t test but not to the one sample t test performed in the substracted images.
I have checked in questions from previous users/glm and randomise webs without success, could anyone assist me with this?
Many thanks,
Jose
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|