Thank you Rich,
My original question:
Is it possible to use multicomp, or something similar, on a multi-level
aov model?
Richard Heiberger's abbreviated answer with my further questions:
R: Yes, it is possible to use multicomp for multi-strata designs.
... respecify the design as a single stratum design and use that.
D: I made it this far and could see where the different RMS and df
values
changed the resultant F values. I was also able to generate the mean
comparisons for each of the strata in my model. But I suspect from the
next part of your answer that I cannot rely on this.
So, how will this impact on the mean comparisons?
R: ... use ... multicomp.default and specify the denominator mean square
yourself.
D: How? I could not decipher this from the help pages, nor source. As
near as
I can figure from your Goicoa example (?"mmc.mean") is that the
"lmat.rows=16:20"
argument to multicomp.mmc does this. If so, how do I determine which
rows
to specify for each of the higher level strata in my model?
Thanks again, DaveT.
*************************************
Silviculture Data Analyst
Ontario Forest Research Institute
Ontario Ministry of Natural Resources
david.john.thompson@ontario.ca
http://ofri.mnr.gov.on.ca
*************************************
|