[R-sig-ME] "Animal Model" plus logistic regression

[Juliet Hannah]

For a continuous trait, the animal model can be fit using the kinship
package. I don't think it works with discrete traits yet. My
understanding is that you need to be able to specify a custom
correlation matrix. Can glmer do that?

On Mon, Jun 7, 2010 at 11:41 AM, Bertrand Servin
<bertrand.servin at toulouse.inra.fr> wrote:
> Dear list,
>
> I tried searching in the archive, but cannot seem to find a relevant
> thread on my problem. I would like to fit a generalized linear
> model (logit link) where the random effects are individuals and
> I estimated (from genetic data) the expected correlation between
> these random effects. This is sometimes called "Animal model" because
> it is quite used in livestock genetics.
>
> This could be a simple dataset:
>
> library(lme4)
>
> y=c(1,1,0,0)
> ID=as.factor(1:4)
> V=matrix(c(1,0.1,0.2,0.3,0.1,1,0,0,0.2,0,1,0,0.3,0,0,1),nrow=4,byrow=T)
>
> The model is just logit(y) ~ mu + e
>
> with Var(e)=V.sigma^2_a+I.sigma^2_e
>
> So V is not to be estimated within the model fit, I just want to
> get an estimate of sigma^2_a, and the model residuals.
>
> I was thinking among these lines:
>
> glmer(y~1|ID,family=binomial,doFit=FALSE)
>
> and then modify something within this model.
> Related question, how do I actually fit the model once
> I made the modifications needed ?
>
> Is this the right way to go ?
>
> thanks for your help
> --
> Bertrand Servin
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>