[R-sig-ME] Estimates for groups within fixed effects & also: enough info for estimating variance? MCMCglmm

[Drager, Andrea Pilar]

Hi all,

I've fitted a model with a binary response at the individual level,  
two continuous fixed effects (measured at the individual level but  
containing info on 23 species), and a single random effect of  
"species". My model is mixing well and the results fit the data, but  
what I am really interested in is getting estimates for the fixed  
effect by species, rather than globally.

It is my understanding (per a 2016 reply to a post to this list:  
parameter estimates for all factor levels MCMCglmm) that by  
suppressing the intercept here, I am estimating the first "level" (one  
species?) for each effect, and that the remaining effects are  
deviations from these levels.  How do I code random slopes and  
intercepts for species within each of the fixed effects, rather than  
deviations from a common distribution?

priorS = list(R = list(V = 1, nu = 0, fix = 1),  G = list(G1 = list(V  
= 1, nu = 0.002)))

smodel <-MCMCglmm(fl~  -1 + Zdbh + Znnd,
                   random = ~ species,
                   family = "categorical", verbose=F, pr = TRUE,  
start=list(QUASI=FALSE),
                   data=IHF,prior=priorS,
                   nitt=500000,burnin=5000,thin=100)


A second part of the question is that I would like to include a random  
effect to measure the influence of phylogenetic autocorrelation. I  
know how to code this using a distance matrix (see below), however, I  
am concerned that statistically, "species" and "phylo" may not have  
enough information to partition the variance in a meaningful way.  
Perhaps running three models (random=species, random=phylo,  
random=species + phylo) and comparing goodness-of-fit could be useful?  
I would greatly appreciate any insight. Many thanks!

prior = list(R = list(V = 1, nu = 0, fix = 1), G = list(G1 = list(V =  
1,nu = 0.002),
                                                           
G2=list(V=1,nu=0.002)))

model <-MCMCglmm(fl ~ Zdbh_mm + Znnd,
                  random = ~species + phylo,
                  family = "categorical", verbose=F, pr = TRUE,  
start=list(QUASI=FALSE),
                  ginverse=list(phylo=inv.phylo$Ainv),data=IHF,prior=prior,
                  nitt=500000,burnin=5000,thin=100)
Andrea Pilar Drager
PhD. student
Ecology and Evolutionary Biology, Rice University