rbartpackages

Python wrappers of R BART (Bayesian Additive Regression Trees) packages.

rbartpackages lets you use a few R BART implementations (including the 3 most popular ones) from Python with a natural Python experience. The Python wrappers closely follow the R packages they wrap, so they should be obvious to use coming from R if one wants to switch to Python.

The packages included are:

The wrappers are incomplete and only wrap the main functions and objects of the packages. The most complete wrapper is dbarts. If you need more functionality wrapped, feel free to open an issue on github. Alternatively, this library provides simplified tools and utilities to let the user wrap arbitrary R classes and functions, see the guide and the reference.

Installation

pip install rbartpackages

You also need R with the latest version of the package(s) you want to use installed (BART, dbarts, bartMachine from CRAN; BART3 from rsparapa/bnptools and missBART from yongchengoh/missBART on GitHub). bartMachine additionally requires Java. If you install polars[pyarrow], dataframes are returned as polars dataframes instead of pandas dataframes. For convenience, you can install both with pip install 'rbartpackages[polars]'. If you install jax (separately or with pip install 'rbartpackages[jax]'), jax arrays are accepted as input in place of numpy arrays, but output arrays remain numpy.

Usage

import numpy as np
from rbartpackages import BART3

x_train = np.random.randn(100, 5)
y_train = x_train[:, 0] + 0.1 * np.random.randn(100)

bart = BART3.gbart(x_train=x_train, y_train=y_train, ndpost=200)
y_pred = bart.predict(x_train)  # shape (ndpost, n)

R argument names with dots are passed with underscores (x.train → x_train).