ROBUST ESTIMATION OF POPULATION-SIZE WHEN CAPTURE PROBABILITIES VARY AMONG ANIMALS

A model is given for multiple recapture studies on closed populations which allows capture probabilities to vary among individuals. The capture probability of each individual is assumed to be constant over time. Based on this model, a nonparameteric estimation procedure is given for population size. The estimator involves selecting one of a sequence of estimators which are each linear combinations of the capture frequencies. The individual estimators are derived from the generalized jackknife method. A goodness of fit test for the model''s assumption that individual capture probabilities do not change during the study is analyzed. The robustness of this estimation procedure is investigated with a simulation study. Robust to moderate variations in individual capture probabilities may occur in commonly used short-term livetrapping studies.