Hi, yes I _think_ your thoughts on the expanded F-tests are correct.
Indeed, I don't think you want to get into subject interactions; if you
did, the analysis would be closer to the final example on the web page I
linked to yesterday, I guess with A and B as fixed effects and subjects as
random effects.
Good luck, Steve.
On Sat, 8 Mar 2003, edward vessel wrote:
> Hi Steve -
>
> ok, now things are much more clear =) I think I know how to proceed
> from here. One last question ... when I expand the design to the full
> 16 subjects, my F statistics will change slightly, right? In other
> words, I will now include the first two columns from _each_ subject in
> my F(A) test (i.e., it will have Subj x A-1 d.f.). Same for F(B) and
> F(AB). Also, I could additionally add an F(Subj.) test that included
> the subject variables only, correct? I don't really see how I would go
> about testing subject x condition interactions, since the analysis you
> mention does not explicitly include and subject interaction terms.
> But, that's ok, I am willing to forego such an analysis =) As for the
> contrasts of interest ... the only additional contrast of interest that
> I really want to look at that you did not include is A1xB1 - A3xB1 ...
> but I think I understand how to create that contrast now.
>
> thanks a bunch!
|