[R] Mixed effects and hierarchical error terms

[Richard Gunton]

Please could someone help me understand the difference between the following two
models?

lme (Y ~ X, random= ~1|G1/G2)
aov (Y ~ X + Error (G1/G2))

Since random effects can be considered as additional error terms, I would have
thought these two model formulations should amount to the same analysis, but
the outputs (using suitable data, with X, G1 and G2 being factors) look very
different.

I note that the lme() model can be fitted even where the levels of X do not vary
within levels of G2, whereas the aov() model output is accompanied by a warning
message ("Error() model is singular...") in this case.

Thanks,

Richard.


--
Richard Gunton
PhD student - Ecology and Evolution group
School of Biology, University of Leeds, LS2 9JT, UK
Room 10.16, Miall Building   Tel: 0113 3432825

http://www.personal.leeds.ac.uk/~bgyrg

~ Opinions expressed in this message are not attributable to the University of
Leeds ~