Hi all,
For a pilot study, we have 10 participants who completed two visits (drug/placebo), and we wanted to test the moderating effect of Trauma history (group variable) on drug (vs. placebo). Rather than use the GLM 2X2 Mixed ANOVA, a colleague suggested to use randomise (which is mentioned in the last sentence of the GLM wiki). After consulting with her and reading through previous FSL questions, I did the following:
1. For each participant, I used fslmaths to subtract Drug - Placebo copes (and vice versa; Placebo - Drug) to obtain difference scores. I used 1stlevel copes from /reg_standard/stats/ .
2. I then merged the 10 participants' difference copes into a single 4D file using fslmerge (I created separate files for Drug-Placebo & Placebo-Drug). Participants were ordered according to Trauma grouping (1 1 1 1 1 0 0 0 0 0 ), where 1 = Trauma, 0 = None.
3. Using FEAT, I created the design file template files (.mat and .con). See attached.
4. Finally, I used randomise to run the Two Sample Unpaired T-Test:
randomise -i $INPUT -o $OUTPUT -d 3rdLev_TwoPairT.mat -t 3rdLev_TwoPairT.con -m mask -T --uncorrp
From the output, it says it ran successfully:
Loading Data:
Data loaded
1 permutations required for exhaustive test of t-test 1
Doing all 1 unique permutations
Starting permutation 1 (Unpermuted data)
Critical Value for: tfce_corrp_tstat1 is: 7544.8
1 permutations required for exhaustive test of t-test 2
Doing all 1 unique permutations
Starting permutation 1 (Unpermuted data)
Critical Value for: tfce_corrp_tstat2 is: 4687.56
Finished, exiting.
Successfully completed.
However, for each t-test, I see only one permutation. I know that randomise can generate multiple permutations (although some may be redundant). Thus, I wanted to ensure my design setup, etc, was correct.
Thanks for any input!
Erin
|