A0384
Title: A Bayesian parametric approach to estimate misidentification errors in capture-recapture
Authors: Davide Di Cecco - Unitelma Sapienza (Italy) [presenting]
Andrea Tancredi - Sapienza University of Rome (Italy)
Abstract: The standard methodologies for capture-recapture and species abundance analysis assume the absence of misidentification errors. However, the presence of such errors is well documented in both contexts. For example, in genotype surveys for animal abundance studies, samples such as fur or faeces are collected in an area and analysed to extract DNA. The occurrence of sequencing errors results in the generation of fictitious genotypes that cannot be linked with other cases and, consequently, leads to erroneous inflation of the number of animals captured once (singletons). In microbial diversity studies, environmental samples, such as soil or water, are analysed to extract DNA or RNA and classify the microbial community into different species. Changes in the species abundance distribution are utilised to assess the impact of factors such as climatic change or the use of chemicals on the health of an ecosystem. Again, sequencing errors lead to the recording of fictitious species, which inflates the number of singletons. A fully Bayesian parametric approach is employed to model spurious singletons as false-negative record linkage errors, and an MCMC algorithm is presented to estimate the posterior distribution of the true number of captures and the number of unsampled units. The MCMC can be specified for different parametric assumptions, and we define exactly the families of distributions for which the algorithm can be adapted.
