Hi,
I am trying to figure out how to specify the model for a repeated
measures experiment using lme. I have the following structure for the
experiment:
There are 3 sites and at each site there are two depths, so site and
depth are crossed. At each site and depth there are 12 fragments of
Coral, each representing a seperate genotype, so genotype is crossed with
site and depth. On each coral fragment the size of 6 coralites is
measured so coralite is nested in fragment. Now in the model genotype
should be a random effect, and we are trying to measure the effect of
depth and site. So I fit the model
mod1<-lme(Factor1~Islands*Depths+Frags:Depths+Frags:Islands +
Frags:Depths:Islands,dhfactan,random=~1|Frags/Corallites)
and
anova(mod1)
numDF denDF F-value p-value
(Intercept) 1 289 0.000000 1.0000
Islands 2 289 2.503565 0.0836
Depths 1 289 10.022984 0.0017
Islands:Depths 2 289 3.772553 0.0241
Depths:Frags 22 289 7.970114 <.0001
Islands:Frags 22 289 7.806390 <.0001
Islands:Depths:Frags 22 289 6.554670 <.0001
Now I know I have done something wrong here becuase there seem to be far
too many degrees of freedom
in the denominator. I am not sure if this is due to how I coded the data
frame or how I specified the model.
The data frame looks like
Corallites Frags Islands Depths Factor1
1 1 1 1 1 0.9476487
2 2 1 1 1 0.4810762
3 3 1 1 1 -0.2544952
4 4 1 1 1 -0.3711468
5 5 1 1 1 -1.0486175
6 6 1 1 1 -0.9323673
7 1 2 1 1 0.6410660
8 2 2 1 1 0.7590714
9 3 2 1 1 2.3403437
10 4 2 1 1 1.8750632
11 5 2 1 1 0.7945611
12 6 2 1 1 2.0533272
Where Factor1 is the response. Any advice is much appreciated.
Thanks
Nicholas