Software & Code

Software

• bayesbridge v0.2 — documentation
  Python package for Bayesian sparse regression based on the Bayesian bridge priors. The package implements the standard (Polya-Gamma augmented) Gibbs sampler as well as the CG-accelerated sampler of Nishimura & Suchard (2023).

• PRSBridge
  Command line tool for constructing polygenic risk scores (PRS) from summary statistics, built on an adaptation of the bayesbridge package. Methodological details and benchmark comparisons against other PRS methods are provided in Dun, Y., Chatterjee, N., Jin, J., & Nishimura, A. (2026). The benchmark results can be reproduced by this code.

• htdg
  R package for sampling from high-dimensional truncated gaussians. In particular, it implements the Hamiltonian zigzag sampler of Nishimura, A., Zhang, Z., and Suchard, M. A. (2025). Benchmark results and example code are provided in this R Journal article.

• CausalSurvival
  R package implementing state-of-the-art doubly robust methods for estimating counterfactual survival curves, including augmented inverse probability weighting and targeted maximum likelihood estimation. More traditional procedures such as inverse probability weighting and stratified Cox model are also included.

Code

• Bouncy Hamiltonian Monte Carlo
  Python module implementing the bouncy Hamiltonian sampler of Chin, A. and Nishimura, A. (2024). The Gibbs samplers using the bouncy sampler for conditional updates, as presented in the article, are implemented here and here.

• Hoseshoe scale sampler
  Efficient rejection sampler for updating the local scale parameter in Gibbs sampling posterior distributions under (regularized) horseshoe models. Details and theoretical analysis can be found in the appendix of Nishimura and Suchard (2022).

• Discontinuous Hamiltonian Monte Carlo
  Python module implementing the discontinuous Hamiltonian Monte Carlo of Nishimura et. al. (2022). Other codes used in the paper are also provided, including the modules to efficiently compute the log-likelihoods and their gradients of the Jolly-Seber and PAC Bayesian inference.

• Probabilistic importance weighted matrix factorization
  Python module for the Bayesian heteroscedastic matrix factorization model as described in Yang et. al. (2017).

• (Recycled) No-U-Turn sampler
  Matlab functions for the No-U-Turn sampler of Hoffman & Gelman (2014) as well as its improvement via the recycling algorithm of Nishimura & Dunson (2016).