Dear FSL users,
I am doing an unpaired t-test with randomise on a data set pooled from
two different studies. I have 16 subjects per group, so 32 subjects in
total pooled from the 2 studies. The stimulus of interest is the same
accross the studies but its timing is different. I would like to add a
covariate to correct for the study effect if any. I am using two
additional columns in the unpaired design with entries zero for study
one and 1 for study two. I am not demeaning. I get different results
when I use instead 1 and -1 as entries instead of zeros. From what I
read on the FSL list it seemed that when looking at the group difference
in an unpaired design the choice of the entries for the covariates
(discussion was on gender) should not matter.
Any help is much appreciated,
Best
Paul
|