[R-sig-ME] GLMM parameter interpretation

[Erik Frenzel]

Adam,
I had a similar question and found this post helpful:

http://r.789695.n4.nabble.com/lme4-and-Variable-level-detection-td881680.html

Erik

> ----------------------------------------------------------------------
>
> Message: 1
> Date: Thu, 19 Jan 2012 23:22:00 -0500
> From: Adam Smith <raptorbio at hotmail.com>
> To: <r-sig-mixed-models at r-project.org>
> Subject: [R-sig-ME] GLMM parameter interpretation
> Message-ID: <BAY170-W226D02A52B38596C044A97A1870 at phx.gbl>
> Content-Type: text/plain; charset="iso-8859-1"
>
>
> All,
>
> A quick question, and an easy one I expect.
>
> Suppose the following generic overdispersed log-linear model:
>
> Count ~ A + B + C + (1|level1) + (1|level2) + (1|obs)
>
> A, B, and C are all factors.
> A does not vary within level1 subjects.
> B does not vary within level2 subjects.
> C varies within level 1 and level 2 subjects.
>
> Does the conditional nature of the GLMM and the lack of variation in A (within level 1 subjects) and B (within level 2 subjects) result in nonsensical or uninterpretable coefficients for these two covariates?
>
> Thanks,
>
> Adam Smith
> Dept. Natural Resources Science
> 105 Coastal Institute in Kingston
> University of Rhode Island