B1679
Title: Species abundance estimation in the presence of record linkage errors: An ABC approach
Authors: Davide Di Cecco - University of Rome La Sapienza (Italy) [presenting]
Andrea Tancredi - Sapienza University of Rome (Italy)
Abstract: The phenomenon of one-inflation in zero-truncated count data has been receiving increasing attention in capture-recapture and species abundance literature. The phenomenon manifests itself as an abundance of singletons (units captured exactly once), which suggests the necessity of explicitly modeling a mechanism for this deviation. We distinguish two possible causes for one-inflation: the erroneous inclusion of spurious units, and missed links in a preliminary record linkage step. Note that we do not have access to the record linkage procedure, but only to the aggregated count data. While the first mechanism can easily be estimated both in frequentist and Bayesian context (via simple Gibbs-based MCMC), we found record linkage errors to be hard to investigate outside a Bayesian ABC approach. As a matter of fact, missing links errors are sometimes tacitly treated as spurious data. We implemented an ABC algorithm for various count distributions and applied it to the estimation of the number of microbial species in literature data.
