Frequentist and Bayesian Inference for Hierarchical Hidden Markov Models Using Stan

Luis Damiano (Department of Statistics, Iowa State University) , Michael Weylandt (Department of Statistics, Rice University) , Brian Peterson (University of Washington) , Michael Weylandt (Department of Statistics, Rice University)

The BayesHMM R package implements a comprehensive framework for Bayesian inference of Hidden Markov Models, including a tailored DSL for model specification, automatic translation to the Stan modeling language, and a variety of diagnostic tools. Under the hood, the package uses the techniques we discuss in our tutorial, but additional performance improvements could be achieved by rewriting the backend to use Stan’s new built-in HMM tools. The package should be stable, but has not been thoroughly validated.

Direct Link:

Package Documentation:

Related Publications: A Tutorial on Hidden Markov Models using Stan