 |
 |
On Wed, Dec 12, 2012 at 5:47 AM, Ben Becker <[log in to unmask]> wrote: > Dear SPMers, > > I want to test the overall significance in my model with an F-test. The model contains a within-subject factor (timepoint: pre- and post-treatment) and a between-subject factor (group: patients and controls). I used the flexible factorial option in spm to define the 2nd level model with the following factors: > 1. subject > 2. group > 3. timepoint > the design matrix with all maineffects has the rows > 1 & 2 group effects > 3 & 4 timepoint > 5 - Inf subjects > > How to define an overall F contrasts to test the overall effect? (1) I'm not sure what you mean by overall effect, nor do I think it would be very useful. It may not even be possible as most statistic books only discuss effects of each factor and different factors have different error terms. (2) The main effect of group -- NOT VALID because you only have within-subject error term (3) The main effect of condition -- VALID (4) The interaction of group*condition -- VALID Hope this helps. > > Thanks in advance & best regards > > Ben