[R] TukeyHSD for repeated measures aov ?

[Dieter Menne]

Sylvain Clément <sylvain.clement <at> univ-lille3.fr> writes:

> I search the archive for finding a simple 
> solution for using TukeyHSD with a multistratum 
> aov result (a repeated emasure anova). The Question have
> been asked but I've found no clear answer.
> 
> res<-aov(y~Fa*Fb+Error(Subject/(Fa*Fb)) )

I would suggest that you use lme in package nlme for models of this type. It's 
much easier to handle, and there is excellent documentation including a book by 
the authors of the package (Pinheiro/Bates).

However, a ready made multiple-testing procedure is not available for lme. 
Package multcomp provides one for lm type models, and Thorsten Hothorn has 
shown how it could be used with glm (and possibly lme in a similar way).
 
http://finzi.psych.upenn.edu/R/Rhelp02a/archive/21117.html 


Dieter