A Bayesian Parametric Approach to Capture–Recapture with Misidentification

The absence of identification errors is a fundamental prerequisite in capture–recapture for population estimation and species abundance estimation. However, such errors have been reported and studied in both contexts. The most common case is the failure to recognize a previously detected entity, i.e., a false negative record linkage error. This results in artificious entities, sometimes referred to as “ghosts”, which typically constitute spurious singletons. We present a Bayesian parametric approach to the problem, which is applicable when data are summarized as number of captures. We develop a Markov chain Monte Carlo algorithm to estimate the proposed model and illustrate the performance of our approach on four datasets available in the microbial diversity literature.