bmkconverge {bmk} | R Documentation |
MCMC chain convergence diagnostic.
bmkconverge(inputlist1, binsize = 1000)
inputlist1 |
A list of the MCMC chains |
binsize |
a scalar giving how large each bin should be for consecutive batches. outputs the Hellinger distances between the sampled distribution for one scenario against the other. |
This takes an MCMC chain and divides it into batches of
size binsize
and calculates the Hellinger distance
between consecutive batches.
Boone EL, Merrick JR and Krachey MJ. A Hellinger
distance approach to MCMC diagnostics. Journal of
Statistical Computation and Simulation,
DOI:10.1080/00949655.2012.729588
.
## Not run: library(dismo); library(MCMCpack); data(Anguilla_train) b0mean <- 0 b0precision <- (1/5)^2 mcmclen = 1000 burn=10000 MCMC.one <- MCMClogit(Angaus ~ SegSumT+DSDist+USNative+as.factor(Method)+DSMaxSlope+USSlope, data=Anguilla_train,burnin=burn, mcmc=mcmclen, beta.start=-1, b0=b0mean, B0=b0precision) ## End(Not run) data(MCMCsamples) mcmclen <- 1000 bmkconverge(MCMC.one,mcmclen/10)