[R-sig-ME] How to assess significance of variance components(Please discard previous e mail... but read this one)
chantepie at mnhn.fr
Thu Nov 15 18:53:03 CET 2012
Thank David for the package reference.
Sorry, I did not recall the context of my work but I am doing animal models on
Poisson distributed traits. So I have to use MCMCglmm package -and
consequently I am in a Bayesian framework).
I want to test an increased in genetic variance (VA) by AGE. The major issue
come from the fact that univariate models using age-classes show a decrease
with VA across age and the random regression show an increase with age. I want
to test the accuracy of AGExVA parameter of the random regression.
-MCMCglmm give posterior distributions well shape for all parameters of each
model, so it cannot be a criterion of rejecting a model.
-According to Jarrod, DIC is not usable with Poisson distribution trait.
I am unable to decide if univariate or RR which provide the best explicable
Have someone an idea to define a criterion?
thank in advance
More information about the R-sig-mixed-models