Estimation of Capture Probabilities Using a GEE Approach

Abstract

Heterogeneity in capture probabilities represents a serious problem when estimating population size in capture-recapture studies. A generalized estimating equation (GEE) approach is applied to data from a capture-recapture experiment to model capture probabilities through a logit-link function of the covariates. The model accounts for heterogeneity resulting from individual characteristics as well as correlation among trapping occasions. In this paper, we build-up a model to use GEE in capture-recapture methodology and argue that heterogeneity and correlation among capture occasions should be accounted for. Quasi-likelihood Information Criterion (QIC) is used for selecting best fitting model. The estimates of capture probabilities may then be used to estimate population size and a real application is revisited for illustrative purposes.