Software
moveHMM
We developed the R package moveHMM for the analysis of movement data with hidden Markov models.
CRAN: description, documentation, archives…
Github: latest (unstable) version of the code.
Get started with the vignette. There, we describe the functionalities of the package in detail, and illustrate their use on elk tracking data.
Short video presentation for MEE.
momentuHMM
The R package momentuHMM extends moveHMM to more general and flexible models. Additional features include: unlimited number of data streams, inclusion of covariates on the observation distribution parameters, centres of attraction, multiple imputation to account for irregular sampling and/or measurement error, etc.
On CRAN.
On Github.
Get started with the vignette.
hmmTMB
hmmTMB is a general R package for hidden Markov models, which uses TMB and mgcv to allow for random effects and non-parametric covariate effects on model parameters (with automatic smoothness selection). The package is available on Github, and it is described in several vignettes:
Overview of hmmTMB workflow (energy price case study)
Other R packages
smoothSDE can be used to fit varying-coefficient stochastic differential equations, described in this preprint.
MScrawl is a package to fit state-switching continuous-time correlated random walks using MCMC, as described in this paper.
localGibbs implements the local Gibbs model for animal movement and habitat selection, presented in this paper and this paper.
Miscellaneous
Vianey Leos Barajas and I wrote a tutorial about using Stan to implement hidden Markov models, in particular to analyse animal movement data. In that document, we re-analyse the wild haggis data set from the moveHMM paper.
A while ago, I also wrote a document about implementing hidden Markov models in R and/or C++ (using Rcpp). It describes two examples in detail: a 2-state HMM with Poisson-distributed observations, and a 3-state HMM inspired by movement models. R code is provided for simulation, estimation, and inference in these models.