[R-sig-ME] R2 measure in mixed models?

[David Duffy]

On Mon, 1 Mar 2010, Juliet Hannah wrote:

> How would you go about summarizing the how much variation a SNP (fixed effect)
> explains (in the mixed model setting)/ or the "strength" of the SNP.

In a way, that is an easy one, as we can plug in our estimates for the 
minor allele frequency and genotypic means into the population genetic 
model for the trait, where our estimates of the total genetic and 
environmenal variances will usually come from other (pedigree based) 
studies.  The contribution of the fixed effect is expressed as a 
proportion of the genetic covariation between related individuals in the 
population (including with themselves).  It seems to me that if one wanted 
to do this all in a single model, the SNP would be included as a random 
effect.  The usual measured genotypes model fitted to pedigree data ends 
up estimating the genotypic means and the residual genetic variance, which 
I don't find completely satisfactory.

It becomes more interesting for categorical traits/GLMMs, where 
some assumptions about gene by gene interaction and the appropriate link 
function have to be made.

Cheers, David.

-- 
| 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