NIMBLE: Hierarchical Statistical Models for MCMC and Programmable Algorithms

A system for writing hierarchical statistical models largely compatible with 'BUGS' and 'JAGS', developed by Perry de Valpine. It includes default methods for MCMC, Laplace Approximation, and Monte Carlo Expectation Maximization, and allows users to write nimbleFunctions to operate models and compile them via custom-generated C++. The system extends the 'BUGS'/'JAGS' language by making it extensible, enabling the addition of new distributions and functions.

Use Cases

• Implementing custom MCMC samplers based on the extensible nimbleFunction system.
• Building hierarchical statistical models based on compatibility with BUGS/JAGS languages.
• Compiling model algorithms to C++ for performance based on the described compilation feature.
• Adding new probability distributions or functions to a modeling language based on the extensibility feature.

Strengths

• System is extensible, allowing new distributions and functions to be added.
• Includes default methods for MCMC, Laplace Approximation, and Monte Carlo Expectation Maximization.
• Provides a full User Manual available at a specified website.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

paperswithcode