[R-sig-ME] binary trait

[Jarrod Hadfield]

Hi Mohamed,

Binary animals models tend to mix poorly but this is quite extreme. From 
the output you have does it look like h2 is very small or very large? 
Also, could you give a quick summary of the data (size, number of farms, 
number of years, many relatives/few relatives ...)

Cheers,

Jarrod





On 01/12/2016 11:00, Mohamed Salem wrote:
> Dears,
> I am trying to use MCMCglmm to estimate heritability for binary trait.
> I used this model
> "
>
> prior <- list(R = list(V = 1, fix = 1), G = list(G1 =
>
> list(V = 1, nu = 1000, alpha.mu = 0, alpha.V = 1)))
>
>
>
> model1 <- MCMCglmm(SB ~ 1 + Farm +year, random = ~animal, family =
> "ordinal",
>
> prior = prior, pedigree = Ped, data = Data, nitt = 1e+06,burnin = 10000,
> thin = 100)
> and when I diagnosed the MCMC work by autocorr.diag(model1$VCV)
> I found this results
>           animal units
> Lag 0    1.0000000   NaN
> Lag 100  0.9786790   NaN
> Lag 500  0.9092071   NaN
> Lag 1000 0.8360430   NaN
> Lag 5000 0.4860764   NaN
>   how can I avoid this problem?


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