[R-sig-ME] error in MCMCglmm (regarding priors)
David Villegas Ríos
chirleu at gmail.com
Thu Oct 15 15:26:34 CEST 2015
Hi all,
I'm running this model with this prior:
prior1=list(R=list(V=diag(4),nu=1.002),G=list(G1=list(V=diag(4),nu=1.002)))
mcmc1=MCMCglmm(cbind(t1,t2,t3,t4)~(trait-1)+trait:season+trait:tl_s+trait:year,
random=~us(trait):id,
rcov=~us(trait):units,
data=widew2t,
family=c("gaussian","gaussian","gaussian","gaussian"),
nitt=100000,thin=100,burnin=10000,
prior=prior1,
pr=TRUE)
And I get two different errors (sometimes one, sometimes the other):
Error in MCMCglmm(cbind(t1,t2,t3,t4) ~ (trait - 1) + :
Mixed model equations singular: use a (stronger) prior
Error in MCMCglmm(cbind(t1,t2,t3,t4) ~ (trait - 1) + :
ill-conditioned cross-product: can't form Cholesky factor
I tried other priors like this one:
prior1b<-list(R=list(V=diag(4), nu=4,
alpha.mu=c(0,0),alpha.V=diag(4)*1000),G=list(G1=list(V=diag(4),
nu=4, alpha.mu=c(0,0),alpha.V=diag(4)*1000)))
But still I get the same error messages. I tried with scaled response
variables and still it doesn't work. The only continuous explanatory
variable is already scaled.
Can anyone please help to specify correct priors for this model? Where does
the problme come from?
Thanks
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list