mcmcr: An R package to manipulate Markov Chain Monte Carlo samples

In mcmcr: Manipulate MCMC Samples

knitr::opts_chunk$set(echo = TRUE, warning = FALSE, messages = FALSE, fig.width = 8, fig.height = 12)

Summary

The results of Bayesian analyses are fully described by the posterior distribution for each term in the model. Bayesian software packages such as BUGS, JAGS and STAN use one or more chains to iteratively draw samples from the posterior distributions using Markov Chain Monte Carlo (MCMC) methods.

While a simple parameter such as an intercept corresponds to a single term, more complex parameters consist of multiple terms with their own inherent dimensionality.

The mcmcr package includes two related classes which preserve the dimensionality of the parameters. The first, mcmcr::mcmcarray, stores the samples in an array where the first dimension is the chains, the second dimension is the iterations and the subsequent dimensions represent the dimensionality of the parameter. The second class, mcmcr::mcmcr, stores multiple mcmcarray objects with the same number of chains and iterations. It was developed to allow a set of dimensionality preserving parameters from a single analysis to be manipulated as a whole.

The package includes functions to:

• coerce to and from other MCMC storing classes such as mcarray and mcmc.list objects;
• assess if an object has converged using rhat and esr (effectively sampling rate);
• extract an object's coef table;
• subset an object by its chains, iterations and/or parameters;
• bind a pair of objects by their chains, iterations, parameters or (parameter) dimensions;
• collapse or split an object's chains;
• mcmc_map over an objects values.

The mcmcr package provides an integrated framework for storing, manipulating and summarising MCMC samples.

library(mcmcr)

mcmcr_example
coef(mcmcr_example)

The software archive is at https://github.com/poissonconsulting/mcmcr.

References

mcmcr documentation built on Sept. 7, 2021, 1:07 a.m.