Hello,
I just want to clarify how to apply non-parametric tests using randomise on my task fMRI data. Let’s say I have two contrasts of task conditions I’m interested;
Contrast 1: condition A > condtion B
Contrast 2: condition B > condition A
After completing the first and the second-level analysis with Feat, I am interested in getting;
1) group mean
2) behavioural covariate of interest (e.g.. performance score), both after adjusting for other covariates of non-interest (eg. age)
In Higher level Feat, I have EV1- group, EV2- performance score (demeaned), EV3- age (demeaned), and set contrasts=2 to look at group mean [1,0,0] and performance score covariate [0,1,0].
If I want to perform the same analysis but with TFCE option in randomise, first I need to create input4D files for each contrast (A>B, B<A) by going to the 2nd level feat and using fslmerge to concatenate COPE images representing each subject (fslmerge -t A_B_input cope1.nii.gz cope2.nii.gz …here cope files representing average over runs for each subject). Am I correct?
Then
randomise -i A_B_input -o A_B_out -d <design.mat> -t <design.con> -m <mask> -n 500 -T
(repeat for B_A_input)
Here, can I simply use design.mat and design.con from the 3rd level Feat to get group mean and behavioural covariate results in one shot? I’m wondering because the GLM user guide describes separate examples for calculating group mean and covariates, and applying -D option only for the latter.
For the mask, I can use mask.nii.gz in my higher-level feat folder, right?
Thank you for your help in clarifying this!
Ami Tsuchida
|