I am using 3dmvm to look at main effects and interaction for 3 factors : group, age and sex with resting state data. I also need to look at correlation of a behavioural data with resting states I wonder if the following GLTs are correct :
glt1: correlation of behav1 difference between con and pat , controlling for age and sex?
glt2: correlation of behav1 with con , controlling for age and sex ?
glt3: main effect of group controlling for age and sex?
3dMVM -prefix mvm_group_age_sex_behav1 -jobs 4
-gltLabel 1 con_vs_pat_behav1 -gltCode 1 'group : 1con -1pat behav1:’
-gltLabel 2 con_corr_behav1 -gltCode 2 'group : 1con behav1:‘’
-gltLabel 3 group_main -gltCode 3 ‘group : 1con -1pat’
Subj group age sex behav1 InputFile
subj001 con old female 1.4 subj001_ses-01_rs.nii
subj002 pat old female 1 subj002_ses-01_rs.nii
subj003 pat yon male 1 subj003_ses-01_rs.nii
subj004 con yon male 2.6 subj004_ses-01_rs.nii
A couple of comments-
- For the following two inferences, put an empty space before the last colon:
-gltLabel 1 con_vs_pat_behav1 -gltCode 1 ‘group : 1con -1pat behav1 :’
-gltLabel 2 con_corr_behav1 -gltCode 2 ‘group : 1*con behav1 :’’ \
- Do you expect the the levels for each of the three factor vary in terms of average “behav1”? Check out the discussion here: https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/statistics/center.html
As I can see in the link it says :
Centering is not necessary if only the covariate effect is of interest. Actually I am more interested to see where in the brain there is more correlation with behav1 within each and also between groups. Does GLT1 and 2 that I have help me with that?
Does GLT1 and 2 that I have help me with that?
Yes, but maybe adding one more would further help:
-gltLabel 3 pat_corr_behav1 -gltCode 3 ‘group : 1*pat behav1 :’’ \
Thanks Gang, Would you please also let me know how to calculate the minimum cluster size which can be reported as significant result?