[R-sig-ME] Random & fixed effects on same level

[Malcolm Fairbrother]

Dear Sebastiaan,

What you're proposing to do is entirely reasonable. In essence, as you say, you're using covariates to explain the variation across groups (or alternatively the clustering within groups) at different levels. Any introduction to multilevel modelling should cover this issue (e.g., http://www.cmm.bristol.ac.uk/learning-training/multilevel-models/index.shtml).

Cheers,
Malcolm


Dr Malcolm Fairbrother
Lecturer in Global Policy and Politics
School of Geographical Sciences
University of Bristol


> Message: 2
> Date: Wed, 20 Jan 2010 17:32:30 +0100
> From: "De Smedt Sebastiaan" <Sebastiaan.DeSmedt at ua.ac.be>
> To: <r-sig-mixed-models at r-project.org>
> Subject: [R-sig-ME] Random & fixed effects on same level
> Message-ID:
> 	<930B1A45F446404FA4D99A46F09209C401540E0A at xmail05.ad.ua.ac.be>
> Content-Type: text/plain
> 
> Dear mailing list,
> 
> 
> 
> 
> 
> I have a model where I use leaf characteristics as response variables,
> and populations and trees (nested in populations) as random effects.
> After quantifying the variation associated with the different grouping
> levels (error/within trees/within populations), I've added fixed
> effects. I have information on population level (annual precipitation,
> soil characteristics ...). By adding explanatory variables on population
> level, can I still include 'population' as a random effect (without
> breaking any statistical rule)? My idea was that, by including a fixed
> effect, the fixed effect partially explains the variation added by a
> grouping variable, but now I'm hesitating a bit... Does anybody have
> some references on this topic?
> 
> 
> 
> 
> 
> Thanks in advance,
> 
> Sebastiaan