> >Is there really a marginally significant B main
> >effect? Here are the cell means and marginal means:
> >
> >Factor A Factor B Marg.Means
> > 1 2
> > ----------------
> >1 | 1.5 3.5 | 2.5
> >2 | 5.5 missing | 5.5
> > ----------------
> > 3.5 3.5
RT> As I have stated before (as has GOULD more eloquently), the
RT> proper analysis is this situation DEPENDS entirely upon what you
RT> believe about the POPULATION INTERACTION between factors A and B.
RT> Under an assumption of NO INTERACTION, then STATISTICA'S test is
RT> biased; under an assumption of a particular interaction (e.g. the
RT> missing cell has a value of 3.5 in the population, then STATISTICA's
RT> test is more efficient. Now, my problem with this is that if I
RT> really believe that 3.5 serves as the POPULATION value for the missing
RT> cell, this indicates that an INTERACTION actually does exist, and any
RT> MAIN EFFECT is quite misleading, in the sense that such an effect is
RT> NOT CONSTANT over levels of the other factor. On these grounds, I find
RT> both tests somewhat suspect; one which presumes no interaction, the
RT> other which essentially assumes an interaction which then leads to
RT> ambiguous interpretation of the MAIN effect.
I think one can make a much stronger statement than the one you did. The
effect of factor B assuming no interaction and only a main effect of A is
biased under Statistica's calculations. Assume the missing cell has a
population mean of 7.5. Now we have two main effects, no interaction,
SAS/SPSS/MINITAB/STATA etc. all estimate the appropriate main effects and sums
of squares, while STATISTICA does not. If as you say the missing cell has a
population mean of 3.5, then an interaction is present and the Statistica test
is more appropriate. So the proper method depends on whether you believe a
main effect of the two factors is present or not, and whether or not an
interaction is present. Since you can't estimate the interaction anyway, but
you included both factors in the experiment because you thought there might be
an effect (that is the usual assumption, isn't it?), I find it makes much more
sense to me to use the test that assumes a main effect is present, and leave
the test assuming the interaction is present for another day and another data
set.
Paige Miller
Paige.Miller@f313.n2613.z1.fidonet.org or kp40.118980@kodako.kodak.com
.. A GOOD friend KEEPS the surplus zucchini!! ___ Blue Wave/QWK v2.12