sparseVCBART

R package for a sparse version of VCBART. Uses Bayesian Additive Regression Trees to estimate covariate effect functions in the linear varying coefficient model, with global–local shrinkage.

Installation

System requirements

• R ≥ 4.1
• A working C/C++ toolchain for compiling R packages:
  • macOS: Xcode Command Line Tools (xcode-select --install)
  • Windows: Rtools (matching your R version)
  • Linux: build essentials (gcc g++ make)

R package dependencies

• Rcpp, RcppArmadillo (C++ interface & linear algebra)

From a local checkout:

# In the package root directory:
R CMD build .
R CMD INSTALL sparseVCBART_*.tar.gz

---

Reproducing the experiments

All scripts live in the codes/ directory. A shared helper file, codes/common_sim.R, contains common simulation settings and utility functions used by multiple scripts. Wrappers for competitor methods are in codes/competitor_wrappers.R.

Simulation 1

• VCBART variants: codes/sim1_VCBART_variants.R
• Other competitors: codes/sim1_others.R

Simulation 2

• VCBART variants: codes/sim2_VCBART_variants.R
• Other competitors: codes/sim2_others.R

Semi-synthetic and real data analyses

The scripts for the applied data analyses are provided below:

• Semi-synthetic Political Science Example: codes/PolScidata.R

• Health and Retirement Study (HRS) Analysis: codes/hrs-analysis.R

• Yeast Transcription Factor (TF) Example: codes/yeast_analysis.R