Is there a way to save the acceptance rate of the MCMC algorithm used by MCMClogit?

Question

The R package MCMCpack offers Bayesian logistic regression through MCMClogit. This function prints the acceptance rate of the MCMC (Metropolis-Hastings) algorithm if verbose=TRUE but it does not seem to return it in the mcmc object. Is there a way to access this information in the object?

To test use:

    library(MCMCpack)
    set.seed(12345) 
    n = 1000 
    x = rnorm(n) 
    y = rbinom(n,1,1/(1+exp(-(1 + x))))
    m = MCMClogit(y ~ x, burnin = 5000, mcmc = 1000,
                         tune = 1.3, B0 = 0, verbose = TRUE)

Which prints an acceptance rate of 0.45533 but I do not find this information in names(m) returning NULL or names(attributes(m)) which returns

[1] "dim"      "mcpar"    "class"    "dimnames" "title"    "y"        "call"

The help file suggests that the coda package allows sumarizing information from mcmc objects (see coda) but searching for 'acceptance' in the pdf does not yield any results.

Answer 1

If you don't like the behavior of capture.output, you can use sink instead, which still returns the results but redirects the console output into a file.

sink(file="test.txt")
m <- MCMClogit(y ~ x, burnin = 5000, mcmc = 1000,
               tune = 1.3, B0 = 0, verbose = TRUE)
sink()

out <- readLines("test.txt")
grep( "acceptance rate for beta was", out, value=T)
# [1] "The Metropolis acceptance rate for beta was 0.45533"

class(m)
# [1] "mcmc"

No need for assign or <<-.

If you want to have it as a numeric value, you can extract it as follows.

ar <- grep( "acceptance rate for beta was", out, value=T)
ar <- as.numeric( gsub("^.* was (.*)", "\\1", ar) )
ar
# [1] 0.45533