Approximately Unbiased Variance Estimation for the Jolly-Seber Mark-Recapture Model: Population Size

Summary

Large sample variances for the Jolly-Seber estimates from a mark-recapture experiment are usually estimated by replacing unknown parameters by their estimates in the variance formulae obtained by the so-called delta method. These variance estimates are subject to large biases for small and moderate sized samples. To overcome this problem a new variance formula that is almost unbiased has been developed for population size estimates.

The new formula does not, unfortunately, overcome the problem of the correlation between estimate and estimated variance which leads to underestimates appearing to be more accurate than they are. Indeed removing the bias in the variance estimate tends to make this problem worse.