For more detailed information and a comprehensivediscussion, please refer to our paper associated with this document,available here: For more detailed information and a comprehensivediscussion, please refer to our paper associated with this document,available here:
https://doi.org/10.52933/jdssv.v4i1.79
bartMan is an R-package for investigating and visualising BayesianAdditive Regression Tree (BART) model fits. We construct conventionalplots to analyze a model’s performance and stability as well as createnew tree-based plots to analyze variable importance, interaction, andtree structure. We employ Value Suppressing Uncertainty Palettes (VSUP)to construct heatmaps that display variable importance and interactionsjointly using color scale to represent posterior uncertainty. Ourvisualizations are designed to work with the most popular BART Rpackages available, namely BART, dbarts, and bartMachine. A practicalexample of the package in use can be found in our detailedvignette.
You can install the released version ofbartMan fromCRAN with:
install.packages("bartMan")And the development version fromGitHub with:
# install.packages("devtools")devtools::install_github("AlanInglis/bartMan")You can then load the package with:
library(bartMan)