Modelling continuous-time capture-recapture data

This one-day, in-person course will look at the basic ideas underpinning likelihood-based statistical methods and models for continuous-time capture-recapture data. The emphasis will be on situations where `recapture' observations can occur at any instant, so that the modelling of the process of the observation times themselves is a necessary part of the analysis. In particular it will include the case of a Markov-modulated Poisson process, where there are two parallel random elements: (1) an underlying rate parameter that evolves in time as a Markov process, e.g. representing a behavioural state; and (2) counts of events, e.g. observations, occurring as a Poisson process conditional on the rate parameter.

The ideas will be illustrated through simulation, likelihood calculation and plotting, and analysis of small-scale datasets, motivated by the monitoring of both wildlife and human populations. Application areas include analysis of camera trap data where captures are triggered by animal movements and thus random,and modelling the disease event dynamics underlying patients' medical claims data. The course will be a roughly even mix of presentation and hands-on practical computing in R.

The course instructor is Professor Paul Blackwell, School of Mathematics and Statistics, University of Sheffield. Professor Blackwell has many years of experience working with continuous time processes, advancing and applying methodology, along with extensive publications in this area. 

Registration details will be available in due course