bmhe: A utility package for post-processing of Bayesian models performed in OpenBUGS or JAGS

Bayesian statistics
R
Author

Gianluca Baio

Published

November 15, 2021

Code


Introduction

This utility package consists of a series of functions that can be used to post-process the output of Bayesian models (obtained using OpenBUGS or JAGS). There are mainly three sets of different functions.

Plotting

  • betaplot Trial-and-error Beta plot (using manipulate)
  • coefplot “Coefplot” for the parameters in the model (using tidyverse)
  • diagplot Specialised diagnostic plots to check convergence and autocorrelation of the MCMC run
  • gammaplot Trial-and-error Gamma plot (using manipulate)
  • posteriorplot Various plots for the posteriors in a ‘bugs’ or ‘jags’ object
  • traceplot Makes a traceplot (eg to visualise MCMC simulations from multiple chains, using tidyverse)

Printing

  • print.bugs Modifies the built-in print method for the R2OpenBUGS package to provide a few more options and standardisation
  • print.rjags Modifies the built-in print method for the R2jags package to provide a few more options and standardisation
  • stats Computes and prints summary statistics for a vector or matrix of simulated values

Utility

  • betaPar Computes the parameters of a Beta distribution so that the mean and standard dev are the input (m,s)
  • betaPar2 Compute the parameters of a Beta distribution, given a prior guess for key parameters. Based on “Bayesian ideas and data analysis”, page 100. Optimisation method to identify the values of (a,b) that give required conditions on the Beta distribution
  • ilogit Computes the inverse logit of a number between \(-\infty\) and \(\infty\)
  • logit Computes the logit of a number
  • lognPar Computes mean and variance of a logNormal distribution so that the parameters on the natural scale are mu and sigma
  • odds2probs Maps from odds to probabilities
  • OR Computes the odds ratio between two probabilities

Installation

The package is only available from its GitHub repository. On Windows machines, you need to install a few dependencies, including Rtools first, e.g. by running

pkgs <- c("MASS", "Rtools", "remotes")
repos <- c("https://cran.rstudio.com", "https://inla.r-inla-download.org/R/stable") 
install.packages(pkgs, repos=repos, dependencies = "Depends")

before installing the package using remotes:

remotes::install_github("giabaio/BCEA", ref="dev")

Under Linux or MacOS, it is sufficient to install the package via remotes:

install.packages("remotes")
remotes::install_github("giabaio/BCEA", ref="dev")

Once installed, the package can be called using the library command

library(bmhe)

and its functions used accordingly.