[R-sig-ME] question on MCMCglmm

[Jarrod Hadfield]

Hi,

The prior for the fixed effects is (multivariate) normal, specified as:

prior=list(B=list(mu=mu, V=V))

where mu are your prior means, and V your prior (co)variances. The  
default is mu=0 and V = I*1e+10: depending on the scale of your data  
this is probably close to uniform.

The random effects also have a (multivariate) normal prior, determined  
by the estimated variances. You place a hyper-prior on these variances  
by specifying the G element in the prior.

Cheers,

Jarrod


Quoting emre karaman <emre_marmara2002 at yahoo.com> on Mon, 26 Mar 2012  
01:33:24 -0700 (PDT):

> Dear Dr.,
> For one-way anova model I assume a uniform distribution for overall  
> mean, that is p(mu)=1, and a normal distribution for random effects.  
> However, prior specification for
> overall mean
> is not clear, at least yet for me. Could you please tell the way to  
> overcome this problem.
> Best Wishes. 
> 	[[alternative HTML version deleted]]
>
>



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