Dear Dr. Mclaren,
Thank you so much for pointing out my mistakes, but I still have some questions concerning this matter and I really appreciate some further insights.
1) I assume if we want to test for main effects of dose, i.e. group effect, the contrast would be
T1D1+T2D1+T3D1 = T1D2+T2D2+T3D2
T1D2+T2D2+T3D2 = T1D3+T2D3+T3D3
T1D3+T2D3+T3D3 = T1D4+T2D4+T3D4
did I understand it correctly?
2) In your example "This is for a design with 18 subjects in group 1, 9 subjects in group 2, 2 group terms and 2 conditions",
One way to setup the flexible factorial model is to use one main factor (subject) and one interaction between group and condition, and thus yields a design matrix containing: subject (27 columns) group*condition (4 columns)
I assume there is an alternative way which uses 3 main factors( subject, condition and group) and one interaction between group and condition and results another design matrix containing: subject (27 columns) group (2 columns) condition (2 columns) and group*condition (4 columns)
I would say both ways seem to be correct but they in fact have different dfs. So the key questions are:
Will they yield identical results if proper contrasts are defined?
If not, is there a way to tell which one is more appropriate under certain circumstance?
3) I had trouble understand the meaning of th contrasts
S1G1C1=[1 zeros(1,26) 1 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0]
S1G1C2=[1 zeros(1,26) 1 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0]
I assume the first 27 columns are subject constants but I could not figure out what the following 23 columns stand for (In fact I was expecting 8 columns at most.)
4) Is it possible to test for subject effects or interaction between subject and condition?
5) I had trouble linking the GLM based ANOVA with the "sum all subjects and average" contrast definition method here. Could you please elaborate a little bit more?
6) Is it possible to actually look at the error terms used in SPM statistics?
Many thanks and best regards,
Sincerely and respectfully yours,
Ce