Species sampling with misidentification: a Bayesian parametric approach

We introduce a new species sampling model to account for species misidentification. The assumption is that each sampled individual or specimen has an unknown probability of being misclassified. Misclassified units constitute fictitious cases that artificially inflate the number of species observed only once. We consider standard parametric models for the true number of occurrences of each species, and present a Gibbs sampler algorithm to perform Bayesian infer- ence. The proposed model is applied to a real-world microbial diversity dataset.