[R-sig-ME] multinominal2 MCMCglmm: predicted estimates are not bounded
j.hadfield at ed.ac.uk
Sun Jul 9 20:38:24 CEST 2017
This behaviour sounds unexpected - can you email me the data and I will
take a look?
On 03/07/2017 18:52, Jesse Delia wrote:
> Dear list,
> I'm having issues predicting and plotting values from a model estimated
> using MCMCglmm, and was hoping I might be able to trouble someone for help?
> My model uses cbind to calculate a proportional response (proportion of an
> egg-clutch w/ dry eggs), which should be bounded 0–1. However, I get really
> large predicted values (+30) and negative values using a multinomial2
> MCMCglmm. I am not sure if this is because of my model specifications, or
> how I am using the predict function. I've tried googling, reading course
> notes, and meeting with stats folks at my U without any luck.
> I have the model results and data sheet saved in an R file, but its too big
> to upload to this list. If you think you can have a look let me know and
> Ill send it along.
> I greatly appreciate any help.
> PhD student
> priorT<-list(R=list(V=1e-10,nu=-1), G=list(G1=list(V=1,nu=1,alpha.mu=0,
> alpha.V=100), G2=list(V=1,nu=1,alpha.mu=0,alpha.V=100)))
> modelT<-MCMCglmm(cbind(dry, clutchsize - dry)~careduration*AVErainS17,
> random= ~species+animal, family= "multinomial2", ginverse=list(animal=
> inv.phylo$Ainv), prior=priorT, data=data, nitt=1000000, burnin=50000, thin
> = 300, pr=TRUE, saveX=TRUE, saveZ= TRUE)
> predictions <- as.data.frame(predict(modelT, interval="confidence", type=
> ggplot(data, aes(x = AVErainS17, y = propdry, color = careduration)) +
> ylab("mortality") + xlab("rain")+
> geom_smooth(data=data, aes(x = AVErainS17, y = predictions[,1]))+
> geom_point(aes(shape=species), size=4)
> [[alternative HTML version deleted]]
> R-sig-mixed-models at r-project.org mailing list
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
More information about the R-sig-mixed-models