[R] Repeated measures on a split-plot experiment

[Douglas Bates]

On Tue, May 13, 2008 at 1:40 PM, James Hudson <jmghudson at gmail.com> wrote:
> Please advise the proper format for coding the random effects of a
> split-plot experiment WITH repeated measures. I have been unable to
> figure out how to use either lme or lmer's "grouping" to properly
> assign the error terms. I am measuring univariate responses.
>
> Here is the set-up:
> WP - whole plot factor
> SP - split-plot factor
> SSP - split-split-plot factor, also the repeated measure (time)
> Plot - replicates of each SSP treatment
>
> How do I combine the split-plot effects with the repeated measures [
> Time|Plot(WP*SP)  ] effect into a random statement?

There has been a long exchange on the R-SIG-Mixed-Models at R-project.org
mailing list over the last few days about a similar coding issue.  You
can check the archives at
https://stat.ethz.ch/pipermail/r-sig-mixed-models/

To be able to answer your question I think we would need more
information about exactly which variances and covariances of random
effects you wish to estimate.  I realize that many software packages
allow you to fit "the" repeated measures model, whether or not it is
appropriate for your data.  In R you have more flexibility but that
means that you must be more specific.