merge_sim {bpr}

Merge Multiple Chains

Description

This function is a method for class poisreg. Merge multiple MCMC chains into a unique chain when sampling with nchains > 1 is used.

Usage

merge_sim(object)

Arguments

object

object of class "poisreg" (usually, the result of a call to sample_bpr), with nchains > 1.

Value

The function returns an object of class poisreg with a single element $sim. The returned chains (elements of sim) are obtained by appending the simulated values of each independent chain, under the assumption that they all have reached the same stationary distribution.

Examples

library(MASS) # load the data set
head(epil)

# Simulate multiple chains by setting nchains > 1
fit4 = sample_bpr( y ~  lbase*trt + lage + V4, data = epil, 
                    iter = 1000, 
                    nchains = 4, thin = 2)
# fit4 contains 4 elements with simulation ($sim, $sim2, $sim3, $sim4)

mcmc_diagnostics(fit4)     
# the Gelman-Rubin diagnostics confirms convergence of the 4 
# independent chains to the same stationary distribution

fit4b = merge_sim(fit4) 
str(fit4b$sim)    
# fit 4b contains only one element $sim, of length 1500 
# (which is the result of concatenating the 4 simulations, after removing the first 25% 
# iterations as burn-in and keeping one iteration every two).