Hi all,
I have an issue I cannot solve myself and hope I can get some pointers here:
I have a RCB design with 10 tree species each replicated 5 times in 5
blocks. Now, I classify all species according to their height in 3 groups
(A, B and C). Now, I will choose 2 species of each group and take my
measurements on 8 individuals per species, making sure to get at least one
individual from each block. That gives me sample size of 48. In the end I
want compare the groups with each other. I am not so much interested in the
differences between the species. Then I take a measurement for each
individual in autumn, winter and spring and call this variable 'season'.
My hypothesis is that the groups are different with respect to the
measurement I took, as well as the season will change the outcome of the
measurement, e.g. the measures will be 'ideal' in autumn, ' very
low' in
winter and 'high' in spring.
My data looks similar to this made up example:
mydat<-data.frame(species=rep(1:6, 3, each=8),
block=rep(sample(1:5,replace=T,8), 18),
individual=rep(1:8,18),
group=rep(as.factor(sample(LETTERS[1:3],48,
replace=T)),3),
season=rep(as.factor(c('autumn','winter','spring')),each=48),
outcome=rep(rnorm(144,10,1)))
Here is the model I fit:
fit<-aov(outcome~group*season+Error(season),data=mydat)
summary(fit)
But I guess this is not 100% correct, since I cannot treat block as a
random factor in aov(), can I? How can I proceed here? I tried to do it
with lmer but I don't really understand what I am doing there. I know how
do add the random term but how can treat the season now as a repeated
measures, so that I get the right total degrees of freedom of 47? Also can
I run TukeyHSD() on a lmer model afterwards since I'd like to compare the
means.
Thanks for any help!!
Sanja
[[alternative HTML version deleted]]
