Thank you Cyril,
So if I find regions with no interaction and for these regions want to examine main effects with a TYPE II SS, is any (or both) of the following procedures proper:
1. Include subject (one regressor/subject), group (one regressor/group) and condition (one regressor/condition) in the design matrix (but no regressors for interactions)
2. Same as 1 but no regressors for subjects.
And additionally, if I want to use TYPE I SS I can analyse the effect of a first factor by including only that factor, and in a new analysis examine the effect of a second factor by interpreting its effect when entered together with the first factor? (I don't see any reason for using TYPE I SS, but just for quriosity...)
Best wishes,
Örjan
|