HI,
I've come across a weird issue in the ICA - could you please explain?
I run ICA on 21 subjects each having 3 run
Created a list of the data (see below) and subsequently run melodic -i /Users/dsoto/Documents/fMRI/Expectation/filelist.txt -o /Users/dsoto/Documents/fMRI/Expectation/Melodica30 -v --nobet --bgthreshold=3 --tr=2.2 --report --mmthresh=0.5 -a concat -d 30;
subj1/run1
subj1/run2
subj1/run3
....
subj21/run1
subj21/run2
subj21/run3
What I find is that the 30 components look different if the list specicies the subject order differently , e.g. the reverse order
subj21/run1
subj21/run2
subj21/run3
....
subj1/run1
subj1/run2
subj1/run3
Components 1 and 2 which explain the most of the variance look very similar then components 3 onwards do differ quite a lot
even though melodic is run on exactly the same datasets, but just the sampling order of the subjects data is varied
Can you please explain this?
Cheers
David
|