rmorsomme/PDSIR: Data-augmentation Marko chain Monte Carlo for Fitting the Stochastic SIR Model to Incidence Counts

Efficient data-augmentation Marko chain Monte Carlo (DA-MCMC) algorithm for exact Bayesian inference under the semi-Markov stochastic susceptible-infectious-removed (SIR) model given discretely observed counts of infections. The novelty of this DA-MCMC algorithm is the *joint* update of the high-dimensional latent data. In a Metropolis-Hastings step, the latent data are jointly proposed from a surrogate process carefully designed to closely resemble the target process and from which we can efficiently generate epidemics consistent with the observed data. This yields a MCMC algorithm that explores the high-dimensional latent space efficiently, mixes significantly better than single-site samplers, and scales to outbreaks with thousands of infections. The package contains data from the 2013-2015 outbreak of Ebola in Western Africa to illustrate the use of the algorithm on real data.

install.packages("remotes")
remotes::install_github("rmorsomme/PDSIR")