[R-sig-ME] correlations between fixed effects: groups or individuals for mixed models?

[David Duffy]

On Fri, 10 Aug 2012, john benson wrote:

> I've been using GAMMs and GLMMs to investigate genetic ancestry in 
> relation to environmental heterogeneity.  My data come from individuals 
> (n = 85) that live in social groups (n = 44) across my study area. 
> Thus, I have included social group as a random effect and investigated 
> the influence of several environmental fixed effects. My question: Given 
> that the random effect of social group should account for 
> non-independence between individuals within groups, it seems like I 
> should only be concerned with correlations between independent variables 
> across social groups (rather than across individuals).  Is it the 
> convention with mixed models such as mine to report a correlation matrix 
> between groups, or do folks generally report a correlation matrix 
> between individuals as one would normally do with a GLM or GAM?  I 
> suppose I could report both, but it seems like the correlations between 
> individuals wouldn't be relevant given the inclusion of the random 
> effect.

Two members per group? Or are they overlapping?  And ancestry is unrelated 
to social grouping?

-- 
| David Duffy (MBBS PhD)                                         ,-_|\
| email: davidD at qimr.edu.au  ph: INT+61+7+3362-0217 fax: -0101  /     *
| Epidemiology Unit, Queensland Institute of Medical Research   \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia  GPG 4D0B994A v