Title: Bayesian capture-recapture data modelling with behavioural effects
Authors: Luca Tardella - Sapienza University of Rome (Italy) [presenting]
Danilo Alunni Fegatelli - Sapienza University of Rome (Italy)
Abstract: In the context of capture-recapture sampling we rely on the generalized linear model framework for modelling behavioural effects by regressing the capture occurrence on previous partial capture histories although shortcuts have been embedded to reduce computational complexity whenever possible. In particular, we extend the modelling ideas of using suitable meaningful summaries of individual previous partial histories. This leads to generalizing the Markov dependence in the presence of a non-linear regression function. Theoretical arguments related to the so-called likelihood failure support the use of a Bayesian approach for the estimation of the unknown population size in the presence of behavioral response to capture. Posterior summaries for inferring the population size within the Bayesian logistic regression is carried out exploiting idea of data augmentation in logistic models using the class of Polya-Gamma distributions hence allowing for a general and flexible computational and modelling framework. Special emphasis will be also given to possible alternative Bayesian computation strategies for model selection. Simulated and real data analysis shows the comparative effectiveness of the proposed inferential approach.