Estimating the size of a population from a single sample: methodology and practical issues

Abstract

In contrast to classical capture recapture methods, the single-sample method of Laska, Meisner, and Siegel (1988) (LMS) enables estimation of the size of a population, N*, on the basis of a single survey. For example, it may be desired to estimate the unduplicated number of individuals served by a mental health center during the last year on the basis of a 1-week sample. The time since each of the sampled individuals last engaged in the activity that defines the population is ascertained. The LMS estimator of N* and its unbiasedness property are motivated in a simple way, and an improved LMS estimator is introduced if additional information is available. An empirical assessment of the procedure is made using mental health service data for which the true population size is known. The performance of the extended LMS estimator is a substantial improvement over the standard LMS estimator.