in order to test differences, I created an fsf file with melodic GUI and I ran it through
a) the melodic gui
b) the command-line feat my.fsf
unfortunately results differs.
/usr/local/fsl/bin/fsl_sub -T 20 -l logs -N feat2_pre /usr/local/fsl/bin/feat /home/data/... -D /home/data/.... -I 1 -prestats
this command produce different output (13731 command-line, 12425 gui)
and consequently also the following
/usr/local/fsl/bin/fsl_sub -T 30 -l logs -N feat3_reg -j 13731 /usr/local/fsl/bin/feat /home/data/.... -D /home/data/ -I 1 -reg
in addition, ICA estimation with symm converge in a different number of steps.
and reconstructed component are almost coincident (although with slightly different explained variance) up to the 17th. then they become quite different.
so. do I really have to run melodic with gui , hence stopping my automatic analysis chain ?
thanks in advance
Alberto
|