multistater: A MULTI-STATistical approach to fitting MULTI-STATE models with R

The multistater package can fit several types of multi-state models, including discrete-state discrete-time Hidden Markov Models using both maximum-likelihood and Bayesian approaches (HMMs) and discrete-state continuous-time state-space models (SSMs). The package was created to specify HMM and SSM designs that can be fit in both the frequentist and Bayesian paradigms within a single package. Our specific application of interest is the analysis of ecological multi-state capture recapture studies, especially acoustic telemetry data of fish movement in riverine systems. However, this package could also be applied to data describing similar multi-state systems including medical data in the form of longitudinal panel studies of disease progression and other forms of ecological capture recapture studies where individuals move between discrete geographical states.

The Bayesian models in this package are written in Stan and called via the cmdstanr interface to R. Additionally, this package uses the contributions of several other R packages that already implement the HMMs and SSMs we describe here, especially: msm. Runtime will vary based on the amount of telemetry data analyzed and settings selected for MCMC analysis. The vignette was specified with settings to make it knit quickly on a standard 2020's laptop computer (several minutes at most).