Abstract
The particle Markov-chain Monte Carlo (PMCMC) method is a powerful tool to efficiently explore high-dimensional parameter space using time-series data. We illustrate an overall picture of PMCMC with minimal but sufficient theoretical background to support the readers in the field of biomedical/health science to apply PMCMC to their studies. Some working examples of PMCMC applied to infectious disease dynamic models are presented with R code.
Highlights
Introduction to particleMarkov-chain Monte Carlo for disease dynamics modellersAkira Endoa,⁎, Edwin van Leeuwenb, Marc Baguelina,b,c a London School of Hygiene & Tropical Medicine, London, United Kingdom b Public Health England, London, United Kingdom c Imperial College, London, United Kingdom ARTICLE INFOKeywords: Particle Markov-chain Monte Carlo State-space models Hidden Markov process Particle filter Sequential Monte Carlo ABSTRACTThe particle Markov-chain Monte Carlo (PMCMC) method is a powerful tool to efficiently explore high-dimensional parameter space using time-series data
The middle panel shows transmissibility over time, with an apparent dip between day 50 and 100. This dip can be confirmed by comparing the transmissibility to the transmissibility at time 0, which shows that between day 50 and 100 the transmissibility is below the starting transmissibility in all cases
PMCMC uses particles to efficiently sample plausible trajectories of hidden state variables to integrate them out, while parameters are explored by the MCMC framework
Summary
In many fields of the applied sciences, inference from time-series data is an important class of problems. Despite potentially high demands for efficient inference tools in time-series analysis, PMCMC has not yet attracted enough attention from practitioners in biomedical/health science fields including biology, ecology, epidemiology and public health. This may be because many of the previous methodological literature on PMCMC required the researcher to have a strong mathematical and statistical background as well as a basic understanding of both MCMC and SMC, which might not always be expected in aforementioned fields (especially in early-career).
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have