Numerical versus asymptotic sequential interval estimation of population sizes.

Abstract

The challenge of estimating a population size (N) is usually faced with the well-established mark-recapture sampling scheme. The basic idea is to tag t items of the population and then observe the number of tagged items appearing in a subsequent random sample. The frequencies of tagged versus non-tagged items are informative to estimate N. For construction of fixed-width and fixed-accuracy confidence intervals for N, this paper compares a numerical non-replacement-sampling sequential design against an asymptotic replacement-sampling sequential design. While the former has shown to perform better in terms of expected sample size, the latter was found superior in terms of robustness with respect to the magnitude of N.