Hi, all.
I'm currently running vertex analysis and randomise for subcortical shape group difference examination.
According to the user's guide, FSL asks me to use GLM design matrices twice each for vertex analysis and randomise with these examples:
first_utils --vertexAnalysis --usebvars -i concatenated_bvars -d design.mat -o output_basename [--useReconNative --useRigidAlign ] [--useReconMNI] [--usePCAfilter -n number_of_modes]
randomise -i con1_dis2_L_Hipp.nii.gz -m con1_dis2_L_Hipp_mask.nii.gz -o con1_dis2_L_Hipp_rand -d con1_dis2.mat -t con1_dis2.con -f con1_dis2.fts --fonly -D -F 3
I'm confused as of what the differences for these GLM matrices are.
I currently have made two different types design matrices: #1. one with no covariates, solely examining the group difference and, #2. another with group effect & covariates. I've made two different matrices because as of now I'm not sure whether these covariates have an effect on the group differences and wanted to be open for possibilities.
With these two matrices am I to run
(1) first_utils with #1 matrix -> randomise with #1 matrix
(2) first_utils with #1 matrix -> randomise with #2 matrix
(3) first_utils with #2 matrix -> randomise with #1 matrix
(4) first_utils with #2 matrix -> randomise with #2 matrix
or am I just to pair first_utils and randomise by using the same type of design matrix (ex. options (1) and (4) listed above)?
Or am I just to observe group differences for first_utils and then observe covariates' effects using randomise?
I think I'm having this confusion because I'm not exactly sure of the role design matrices play for first_utils and randomise independently.
Any advice on how to apply these design matrices for these two commands would be immensely helpful.
Thanks so much. Have a great day!
Sincerely, Sooyun Cho.
|