Heterogeneity and finite mixtures in capture-recapture models
by
Shirley Pledger
Victoria University of Wellington

Capture-recapture models have made enormous progress in the last ten years. Heterogeneity between different animals' capture rates often occurs, possibly caused by intrinsic differences between the behaviour of the different animals or by the spatial relationship between traps and home ranges. This heterogeneity can cause substantial bias in the population estimates. Traditionally this problem has been handled by jackknife methods or by moment-based sample coverage estimators or estimating equations.

It is now possible to use finite mixtures to model the heterogeneity, giving unified, likelihood-based models and the resulting model selection precedures, maximum likelihood estimates, and likelihood confidence intervals. The fitting and comparison of these models for a closed population will be shown, using a skink data set from North Brother Island in the Marlborough Sounds.

Some indication will be given of other potential uses of finite mixtures in capture-recapture models.

Date received: August 28, 2001