[R-sig-ME] how to specify priors in MCMCglmm?

Darren Norris doon75 at hotmail.com
Thu May 7 02:33:18 CEST 2009

Dear all,

I would like to know how to specify priors for a MCMCglmm model with a 
binary response. I have read the help (?MCMCglmm) and searched the best 
I can through nabble R and this list.
Unfortunately if the answer is there I am too stupid to understand it.

>From the MCMCglmm tutorial I see how to code a univariate version - but 
then I get lost about how to extend the prior specification to more 
complex models.
I also do no understand the prior terminology or how / why to specify what I 
need -  so if at all possible I need an idiots guide how to specify the 

I work for a conservation NGO and unfortunately we're not affiliated 
with an academic institution so don't have access to books or journal 
articles or any academic statistical support.
Any guidance would be much appreciated.
Many thanks,

R version 2.8.1 (2008-12-22) 

LC_COLLATE=English_United Kingdom.1252;LC_CTYPE=English_United Kingdom.1252;LC_MONETARY=English_United Kingdom.1252;LC_NUMERIC=C;LC_TIME=English_United Kingdom.1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] MCMCglmm_1.09      gtools_2.5.0       combinat_0.0-6     orthopolynom_1.0-1 polynom_1.3-4      pscl_1.03         
 [7] mvtnorm_0.9-4      ape_2.2-3          coda_0.13-4        Matrix_0.999375-18 lattice_0.17-17    tensorA_0.31      
[13] corpcor_1.5.2      MASS_7.2-45       

loaded via a namespace (and not attached):
[1] gee_4.13-13 grid_2.8.1  nlme_3.1-89

#make the data frame- using "sample" so data is not the same but the 
important difference is between probability of events in "cD" and the 
other two ("aD" and "bD")
# we coducted the work in 3 different places, each place has 3 habitat 
#Each habitat type has 2 groups of samples (3 samples per group) each 
group in a habitat has a different type of  bait.
#I am trying to model how the occurence of events was influenced by 
habitat and bait. We checked samples repeatedly (atime).
#Specifying atime and agroup as random effects to deal with potential 
pseudo replication.


#-using prior code examples from the MCMCglmm tutorial (section 1.2: 
produced April 17 2009), but priors are not right.
prior = list(R = list(V = 1, n = 0, fix = 1), G = list(G1 = list(V = 1,n 
= 0)));
MMCMm1 <- MCMCglmm(aevent ~ as.factor(ahabitat), random = ~agroup, 
family = "categorical",prior=prior,data = MyDf, verbose = FALSE)

# I would like to specify, but don't know how to specify priors:
MMCMm2 <- MCMCglmm(aevent ~ as.factor(ahabitat)*as.factor(bait), random 
= ~agroup + atime, family = "categorical",prior=prior,data = MyDf, 
verbose = FALSE)

More information about the R-sig-mixed-models mailing list