mcmc_diagnostics {bpr}

MCMC Convergence Diagnostics

Description

This function is a method for class poisreg. It prints convergence diagnostics and accuracy statistics of the MCMC output.

Usage

mcmc_diagnostics(object)

Arguments

object

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

Details

The printed output of mcmc_diagnostics summarizes some common convergence diagnostics for Markov chains. The first part recaps the total length, burn-in and thinning used for the simulation.

The second part is a table with diagnostic statistics about each chain of the regression parameters. The first column is the effective sample size computed after removing the burn-in and thinning. The last two columns report the value and observed p-value of the Geweke test of equality of the first and last part of the chain.

The last part is printed only if multiple chains are computed. In this case, it reports the Gelman-Rubin statistics to test convergence to the same stationary distribution. Values much larger than 1 suggest lack of convergence to a common distribution.

Value

mcmc_diagnostics returns a list with elements:

chain_length : total length of the MCMC chains.

len_burnin : the length of the burn-in used to compute the estimates.

thin : the thinning frequency used (from object).

effSize : effective sample size of each parameter chain after removing burn-in and thinning. See effectiveSize.

geweke : Geweke diagnostics of convergence of the chains (value of the test and p-value). See geweke.diag

gelman_rubin : if nchains > 1, Gelman-Rubin diagnostics of convergence. See gelman.diag.

See Also

summary.poisreg , plot.poisreg , merge_sim , effectiveSize , geweke.diag , gelman.diag

Examples

# For examples see example(sample_bpr)