Hello,
I'm implementing a 3 grp, heterogeneous slopes model in randomise, and I
was wondering if there are any caveats or areas of caution when doing
permutation testing with such a model?
A graphic of my model is attached. It includes an EV (constant term)
for each of the 3 groups (grp1,grp2,grp3), gender (as a nuisance
variable), and then 3 group*variable regression terms
(vgrp1,vgrp2,vgrp3). The contrasts that I'm interested in include the
positive and negative "main" effect of slope (i.e., average of the 3
separate group slopes), positive and negative contrasts on the
difference of the slope of pair-wise groups, and the F-test looking for
an overall slope difference between any of the 3 groups.
I'm not using any exchangeability block labels file.
My understanding from the web site description is that randomise
automatically figures out the "nuisance" terms in the model from the
contrast vectors. So, I think that running such a model in randomise is
fine, but I suspect that this type of model doesn't get used in
randomise all that often. Thus, I'd appreciate it if one of the
randomise experts would give it their blessing!
BTW: Does anyone have an electronic copy of the Freedman and Lane (1983)
paper that describes the method randomise uses?
thanks,
-MH
--
Michael Harms, Ph.D.
--------------------------------------------------------------------
Conte Center for the Neuroscience of Mental Disorders
Washington University School of Medicine
Department of Psychiatry, Box 8134
Renard Hospital, Room 6604 Tel: 314-747-6173
660 South Euclid Ave. Fax: 314-747-2182
St. Louis, MO 63110 Email: [log in to unmask]
--------------------------------------------------------------------
|