Bioequivalence and Bioavailability Forum

Dear all,

sorry if this is borderline hijacking of the thread, but I read with great interest the thread referenced, where it was proposed to include group as a fixed factor and evaluate Treatment*Group.
When we test for sequence effects we usually apply the between-subject error in the denominator for the F-statistic; the curent logic goes something like this: The sequence effect is a between-subject factor, see e.g. the recent Bogus thread, therefore between-subject variability should be used as error.
So, in all innocence, and given the facts that I am not a statistician and my brain is the size of a walnut, and since I like the logic about testing the sequence effects, let me ask: Which error term is appropriate for Treatment*Group?
Treatment is a within-factor while group - if I get the issue right here - is a between-factor.
(I could of course also ask something like "If we use the power to know in combination with the bogus statement, which error term does the almighty power, in whose hands many statisticians blindly command their spirits, use for the interaction?" - but of course I will not ask that way because it could lead to false assumptions about my opinion about said stats package.)


EM.