I have a dataset with 50 subjects and two timepoints. I would like to see the effect of TP1-TP2, corrected for the covariates age, sex, scanner. I have run the paired_diff analysis with my longitudinal processed data. My second fsgd file now likes this:
GroupDescriptorFile 1
Title Pre_Post_KM_with_sex_age_scanner_as_CV
Class 0_1
Class 0_2
Class 0_3
Class 1_1
Class 1_2
Class 1_3
Variables age
002pair 0_2 18
004pair 0_3 19
005pair 0_1 19
008pair 0_0 21
010pair 0_2 23
012pair 1_1 24
013pair 0_2 24
015pair 1_2 27
016pair 1_1 27
018pair 1_1 28
019pair 0_2 28
022pair 1_1 32
023pair 0_1 32
024pair 0_3 33
0_1 means female sex on scanner 1. 1_2 means male sex on scanner 3 and so on...
So this means I have 6 intercepts and 6 slopes, and my contrast file has 12 inputs. If I just want the effect between timepoint 1 and timepoint 2, controlling for class and age, how should my contrast file look like? If I use 1.6 1.6 1.6 1.6 1.6 1.6 0 0 0 0
0 0 I get no significant clusters which is unlikely when I compare the tabulary data (aseg2stats e.g.)
My command for mri_glmfit was: mri_glmfit --glmdir lh.paired-diff --y mgh_files/lh.paired-diff.volume.mgh --fsgd paired_diff_with_CV.fsgd --C mean_with_cv.mtx --surface fsaverage lh --nii.gz
To display the results it was: tksurferfv fsaverage lh inflated -aparc -overlay /data/Aster_H/Daten/Freesurfer_Longitudinal/Schreglman_S/fsgd/lh.paired-diff/mean_with_cv/sig.nii.gz -fminmax 2 3