[R-sig-ME] lme: random effect nested within fixed effect

[Kingsford Jones]

On Tue, Apr 6, 2010 at 3:37 PM, Andrew Dolman <andydolman at gmail.com> wrote:
[snip]
>
> This is not entirely correct. The syntax will work because the populations
> are uniquely labeled but it does not model population as nested within type.
> It models there being 12 populations (4 for each of the three types).
>
> These two models are not the same:
>
> lme(growth ~ type, random = ~ 1 | population) - with populations uniquely
> identified (A1, A2 ... B1) is not the same as
>
> lme(growth ~ type, random = ~ 1 | type/population) - with populations
> "nested" within type
>

~1|type/population is shorthand for wanting both ~1|type and
~1|type:population.  Since the OP isn't interested in ~1|type, and the
design indicates that ~1|type:population is the same as ~1|population
when the populations are uniquely labelled, the first of the two
models above seems correct.  The tricky part, of course, will be
determining denominator df for the type effects.


Kingsford


>
>
> Andy.
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>