Hello Gang:
Yes, there is only one observation per cell (n=1). So, it is not possible
to perform an exact test in those cases where the denominator of the
corresponding F-statistic is MSE.
However, in most cases of practical interest, e.g.,testing for factor A main
effect, factor B main effect, and AxB interaction, the denominator of the
F-statistic is NOT MSE. For example, to test for factor A main effect:
F(A) = MSA / MSC(A), where MSC(A) = SSC(A)/ a(c-1).
Now, in this particular case, a = 2, and c = 14. Hence, there would be
a(c-1) = 26 dof.
Please see Tables 3 and 4 (p.37) of the documentation in file 3dANOVA.ps for
more details.
Doug Ward
