 |
 |
Dear all,
I am interested in seeing where there is any significant correlation between the within-subject difference under two conditions and an outward measurement.
My first thought would be to build 2nd level a pair-t test model ( to model the within-subject difference) and add the outward measurement as a covariate.
Thus the contrast ( I assume) becomes [1 -1 1] (for positive correlation) or [1 -1 -1] (for negative correlation).
But on second thought, taking the difference imgs of individual subjects and build a multiple regression model with the outward measurement as a covariate also seems valid to me.
So I am a little bit confused here. Are both modelling ways equivalent to each other? Moreover, how does one enter the correct covariate vector into SPM? Since one measurement value corresponds to one subject, in a paired-t test, I assume the design matrix is sth like:
Cond1 Cond2 Sub1 Sub2 Cov
1 0 1 0 s1
0 1 1 0 s1
1 0 0 1 s2
0 1 0 1 s2
....
Did I understand it correctly?
Many thanks and best regards,
Ce