CmdStan User’s Guide
This document is a user’s guide for CmdStan, the command-line interface to the Stan statistical modeling language. CmdStan provides the tools to compile a statistical model written in the Stan probabilistic programming language into a C++ executable program which can then be run to either: do inference on data, producing an estimate of the posterior; generate new quantities of interest from an existing estimate; or generate data from the model according to a given set of parameters
CmdStan provides the programs and tools to compile Stan programs into C++ executables that can be run directly from the command line, together with a few utilities to check and summarize the resulting outputs.
The packages CmdStanR and CmdStanPy which provide interfaces to CmdStan from R and Python, respectively, similarly, JuliaStan also interfaces with CmdStan.
Stan home page
For links to up-to-date code, examples, manuals, bug reports, feature requests, and everything else Stan related, see the Stan home page:
CmdStan, Stan, and the Stan Math Library are licensed under the new BSD license (3-clause). See the Stan Reference Manual Licenses section for licensing terms for Stan and the dependent packages Boost, Eigen, Sundials, and Intel TBB.
Stan documentation: user’s guide and reference manuals
The Stan user’s guide provides example models and programming techniques for coding statistical models in Stan. It also serves as an example-driven introduction to Bayesian modeling and inference:
Stan’s modeling language is shared across all of its interfaces. The Stan Language Reference Manual provides a concise definition of the language syntax for all elements in the language.
The Stan Functions Reference provides definitions and examples for all the functions defined in the Stan math library and available in the Stan programming language, including all probability distributions.
Benefits of CmdStan
With every new Stan release, there is a corresponding CmdStan release, therefore CmdStan provides access to the latest version of Stan, and can be used to run the development version of Stan as well.
Of the Stan interfaces, CmdStan has the lightest memory footprint, therefore it can fit larger and more complex models. It has has the fewest dependencies, which makes it easier to run in limited environments such as clusters.
The output generated is in CSV format and can be post-processed using other Stan interfaces or general tools.