Semi-Markov Arnason-Schwarz models

摘要: We consider multi-state capture-recapture-recovery data where observed individuals are recorded in a set of possible discrete states. Traditionally, the Arnason-Schwarz model has been fitted to such state process is modeled as first-order Markov chain, though second-order models have also proposed and data. However, low-order may not accurately represent underlying biology. For example, specifying (time-independent) involves assumption that dwell time each (i.e., duration stay given state) geometric distribution, hence modal one. Specifying time-dependent or higher-order processes provides additional flexibility, but at expense potentially significant number parameters. extend by semi-Markov for process, dwell-time distribution specified more generally, using, shifted Poisson negative binomial distribution. A expansion technique applied order resulting terms simpler computationally tractable hidden model. Semi-Markov come with only very modest increase parameters, yet permit significantly flexible process. Model selection can be performed using standard procedures, particular via use information criteria. The approach allows important biological inference drawn on times spent different feasibility demonstrated simulation study, before being real corresponding house finches states correspond presence absence conjunctivitis.