Population size estimation with capture-recapture in presence of individual misidentification and low recapture

arXiv (Cornell University)(2023)

引用 0|浏览8
暂无评分
摘要
While non-invasive sampling is more and more commonly used in capture-recapture (CR) experiments, it carries a higher risk of misidentifications than direct observations. As a consequence, one must screen the data to retain only the reliable data before applying a classical CR model. This procedure is unacceptable when too few data would remain. Models able to deal with misidentifications have been proposed but are barely used. Three objectives are pursued in this paper. First, we present the Latent Multinomial Model of Link et al. (2010) where estimates of the model are obtained from a Monte Carlo Markov Chain (MCMC). Second we show the impact of the use of an informative prior over the estimations when the capture rate is low. Finally we extend the model to the multistate paradigm as an example of its flexibility. We showed that, without prior information, with capture rate at 0.2 or lower, parameters of the model are difficult to estimate i.e. either the MCMC does not converge or the estimates are biased. In that case, we show that adding an informative prior on the identification probability solves the identifiability problem of the model and allow for convergence. It also allows for good quality estimates of population size, although when the capture rate is 0.1 it underestimates it of about 10%. A similar approach on the multistate extension show good quality estimates of the population size and transition probabilities with a capture rate of 0.3 or more.
更多
查看译文
关键词
population,estimation,individual misidentification,capture-recapture
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要