Hi,
would would be considered most appropriate to account for some scanner / acquisition protocoll differences?
Data (large N, i.e. many subjects) have been collected from two scanners using two different coils with a single paradigm. The aim is to test for a mean effect of the paradigm across the two scanners and coils.
I thought to add two EVs to the higher level analysis (i.e. entering 1's for the second scanner and the second coil, respectively). On the other hand, the 2 scanners / 2 coils could be modelled as 4 different groups. However, these would vary in their numbers of subjects (with few subjects in some) and I am interested only the the mean effects across groups.
Any thoughts?
Thanks a lot,
Andreas
|