[R-sig-ME] Combining MCMCglmm estimates
Davina.Hill at wits.ac.za
Tue Oct 2 10:06:09 CEST 2012
Dear list members,
I am using MCMCglmm to investigate the effects of an animal’s body mass, age and group size on its reproductive tactic (a factor with four levels), with individual identity and year in the model as random factors. Some values of body mass and age were missing, so I used Amelia to impute ten datasets and then ran the following model on each dataset separately:
k <- length(levels(ds1$tactic))
I <- diag(k-1)
J <- matrix(rep(1, (k-1)^2), c(k-1, k-1))
IJ <- (1/4) * (I+J)
prior = list(R = list(V = IJ, n=4, fix=1),G = list(G1 = list(V = IJ,n = 4), G2 = list(V = IJ,n=4)))
mousemodel <- MCMCglmm(tactic ~ trait-1+mass+age+groupsize, random = ~idh(trait):id + idh(trait):year, rcov = ~idh(trait):units, family = "categorical", burnin = 2000000, nitt = 3000000, thin = 10, data = ds1, prior = prior, verbose = FALSE)
I would like to combine the ten sets of MCMCglmm estimates to get a single inference. The most promising solution I’ve found so far is the mi.inference function in the mix package, which uses Rubin’s rule to combine estimates and SEs from a number of datasets. I don’t know if the rule applies to MCMC objects and I was wondering if someone could advise me whether what I’ve done is appropriate. I extracted the posterior mean and Time-series SE from summary(mousemodel$Sol) for each dataset and ran:
I used ‘Time-series’ SE but I’m unclear what the difference is between this and ‘Naïve’ SE and couldn’t find a definition in the MCMCglmm literature. I'd also like to obtain a single set of HPD intervals if this is possible.
Any help would be much appreciated.
<html><p><font face = "verdana" size = "0.8" color = "navy">This communication is intended for the addressee only. It is confidential. If you have received this communication in error, please notify us immediately and destroy the original message. You may not copy or disseminate this communication without the permission of the University. Only authorized signatories are competent to enter into agreements on behalf of the University and recipients are thus advised that the content of this message may not be legally binding on the University and may contain the personal views and opinions of the author, which are not necessarily the views and opinions of The University of the Witwatersrand, Johannesburg. All agreements between the University and outsiders are subject to South African Law unless the University agrees in writing to the contrary.</font></p></html>
More information about the R-sig-mixed-models