[R-sig-ME] random effect nested in the fixed effect

[Hans Ekbrand]

On Fri, Aug 17, 2012 at 09:39:48PM -0400, li li wrote:
> Dear all,
>     I am starting to use R for mixed models.
>     For example, for the date below,
>       I want to fit the model values=sample+ind(sample).
> Here "sample" is a fixed effect and "ind"
> should be a random effect nested in "sample".

[...]

>    Here is the data structure:
> 
> 
>   values ind sample
> 1  0.03325   1      1
> 2  0.03305   1      1
> 3  0.03185   1      1
> 4  0.03515   1      1
> 5  0.03375   1      1
> 6  0.01180   1      2
> 7  0.01850   1      3
> 8  0.02915   1      4
> 9  0.06200   1      5
> 10 0.03230   2      1
> 11 0.03345   2      1
> 12 0.03385   2      1
> 13 0.03605   2      1
> 14 0.03225   2      1
> 15 0.01145   2      2
> 16 0.01805   2      3
> 17 0.02950   2      4
> 18 0.05995   2      5

This looks like crossed data. It this really is nested data, then it
is implicitly nested as described in
http://lme4.r-forge.r-project.org/book/Ch2.pdf, page 39, and my
formula will not work. For my formula to work, you have to make
samples clearly unique (see page 40 in
http://lme4.r-forge.r-project.org/book/Ch2.pdf):

-- 
Hans Ekbrand (http://sociologi.cjb.net) <hans at sociologi.cjb.net>