Finite mixture models of heterogeneous capture probabilities for mark-recapture estimation of closed population size

摘要

Heterogeneity in capture probabilities among animals is a common problem for estimation of animal population size from mark-recapture data. We model animals as belonging to discrete groups in which animals have the same probabilities of first capture and recapture. For removal data, population size, probabilities of first capture, and mixture proportions are estimated by maximum likelihood for a geometric finite mixture model. For mark-recapture data, a binomial finite mixture for recaptures is combined with a geometric finite mixture for first captures to better estimate mixture proportions. This model can be restricted for the assumption of no behavioral response to first capture.;On Carother\u27s (1973) taxi cab data, estimation with a 2-group mark-recapture finite mixture provided a population size estimate, N = 420, that exactly matched the registered number of cabs. On meadow vole data, estimation with a 3-group model showed heterogeneity in behavioral response to first capture. Simulations show that our 2-group mark-recapture finite mixture estimator with restriction for no behavioral response to first capture is more efficient than Burham and Overtons\u27 (1978) jackknife estimator when the smallest probability of first capture is 0.1 and the number of sampling occasions is 10.