[R-sig-ME] MCMCglmm with datasets of different lengths

[David Duffy]

On Thu, 2 May 2013, Ingleby, Fiona wrote:

> Thanks for the reply, David. I don't think I explained myself clearly 
> enough as I don't think the model with the five traits as response 
> variables would give me what I'm looking for. I want to include the size 
> measurements as a covariate

If the five trait model works, I think you could extract all the 
coefficients you are interested in from the genetic and environmental 
covariance matrices (eg partial out size from G and E respectively).  If 
there is a genetic correlation between size and T1-4, regressing it out 
can be misleading if genetic effects are your main interest.  To fit 
phenotypic causative pathways in a full model probably requires something 
like OpenMx, where you can specify it as a path model.