RStan

rstan is the R interface to Stan, a probabilistic programming language for Bayesian inference. It provides bindings that let users write Bayesian models in Stan’s modeling language, compile them, and run Markov chain Monte Carlo (MCMC) sampling from within R. Stan itself uses a high-performance C++ backend for efficient computation and supports algorithms such as Hamiltonian Monte Carlo and the No-U-Turn Sampler.

The typical workflow in rstan begins with defining a Stan model in the Stan language, either in an external .stan file or as a string in R. The data required by the model is provided as a list in R, matching the data block names in the Stan program. The model can be compiled with stan_model or, in a single step, with stan. Once compiled, sampling is run to generate draws from the posterior distribution. The resulting object, usually of class stanfit, contains the posterior samples and is accompanied by functions to summarize, print, and extract data from the fit.

Key features of rstan include support for multiple chains and parallel computation, automatic differentiation for model gradients, and the ability to specify priors, likelihoods, and transformed parameters within Stan. Users can perform posterior summaries, diagnostic checks, and posterior predictive checks using the fit object and standard R tools. rstan integrates with the broader Stan ecosystem, including the Stan math library and ecosystem packages that provide higher-level modeling interfaces.

rstan is widely used as a foundation for Bayesian modeling in R, and it serves as the backend for higher-level packages such as rstanarm and brms. It requires a working C++ toolchain for compilation and is available on major platforms, with ongoing development to support newer Stan features and language updates.