Dear all,
I am now running TBSS to test the correlation between fiber tracking and language ability in Chinese subjects.
Overall, we have 24 subjects, and all the process went smoothly.
However, later we decide to exclude 5 subjects (leaving 19 subjects) based on other test performance. And I do my analysis
like this:
I didn't redo all the pre-processing steps.
I use fslsplit to split the all_FA_skeletonise.nii.gz, and delete 5 subjects' data, and use fslmerge -t to create a new all_FA_skeletonise.nii.gz to just include the 19 subjects.
and then i use randomize to do the permutation, it went smoothly to do the first 500 permutation, but the second permutation it died at the second permutation stage.
the design is very simple, the first EV is the mean of language ability (1 for each subjects), and the second EV is the demeaned language ability. the first contrast is the mean, and second is the language ability.
here is my question: if we want to exclude some subjects after we finish the preprocessing, is it right just exclude certain images like i did above? if not, how should we do? shall we need to redo all the preprocessing again using new subject list?
thanks in advance,
-Qinghua
Windows Live Hotmail: Your friends can get your Facebook updates, right from HotmailŪ.
