[R-sig-ME] Parameter-expanded priors with MCMCglmm

[W Robert Long]

Hi Jarrod,

This has worked now, and gave results close to what I was expecting.

Thank you very much

Robert Long
PhD Student
University of Leeds


On 07/10/2012 11:56, Jarrod Hadfield wrote:
> Hi,
>
> You need alpha.mu = rep(0,7) too. Also, note that to obtain marginal
> priors on each variance that are scaled F(1,1) you need to have nu=7.
> What effect this has on the covariances, correlations etc I'm not sure.
> I haven't come across any work looking at the prior properties of
> multi-parameter expansion.
>
> Cheers,
>
> Jarrod
>
>
> Quoting W Robert Long <longrob604 at gmail.com> on Sun, 07 Oct 2012
> 11:35:12 +0100:
>
>> I would be grateful if anyone can advise me on how to use parameter
>> expanded priors in the following context.
>>
>> With random=~idh(1+year):hospital I initially had:
>>
>> prior3b <- list(R = list(V = 1, fix = 1), G=list(G1=list(V=diag(7),
>> nu=1)))
>>
>> which worked OK, but I would like to use parameter expanded prior. I
>> have tried this:
>>
>> prior3a <- list(R = list(V = 1, fix = 1), G = list(G1 =
>> list(V=diag(7), nu = 1, alpha.mu = 0, alpha.V = diag(25^2,7,7)))
>>
>> But this gives the error:
>> alpha.V is the wrong dimension for some prior$G/prior$R elements
>>
>> Thanks
>> Robert Long
>> PhD Student
>> University of Leeds
>>
>> _______________________________________________
>> R-sig-mixed-models at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>
>>
>
>
>