[R-sig-ME] Default priors in MCMCglmm()

[Jarrod Hadfield]

Hi,

The default priors have n=0. To have non-informative priors in a
univariate (one response variable) model you need to specify n=-2 and
V=0.00000001.  V should really be zero but it will return an error
when it tries to invert it.  Improper priors can lead to numerical
problems, but hopefully MCMCglmm will pick up on these before they
create problems. Sometimes it doesn't, and you get a variance
"trapped" at zero or a correlation "trapped" at -1 or 1.

Please move on to version 1.09 if you haven't already.

Cheers,

Jarrod



Quoting Christopher David Desjardins <desja004 at umn.edu>:

> Hi I was curious what the default priors are that are used in
> MCMCglmm()? Are they flat unless specified?
> Cheers,
> Chris
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>



-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.