Larson A, Stevens A, Wardlaw G
Tropical Health Program, University of Queensland Medical School, Herston.
Soc Sci Med. 1994 Sep;39(6):823-31. doi: 10.1016/0277-9536(94)90044-2.
This paper describes a method of estimating the size of a population based on matching the individuals from the population appearing in two or more nonrandom samples. The method has been developed by population biologists and demographers who have shown that the simple estimates based on only two samples are very sensitive to the assumption that the two samples are independent. This assumption is rarely met with human populations. However, estimates based on three or more samples can be derived by fitting a log-linear model to a 2k contingency table. The method is applied to a data set of heroin users seeking treatment or counselling or arrested on drug charges in 1988 or 1989 in a small Australian city. The method produced a stable estimate of 1250 heroin users. This includes only users who are vulnerable to arrest and/or interested in seeking treatment. Although estimates produced by this method should be validated with other techniques it provides a simple, quick method to estimate the numbers of people in other hidden populations which are difficult to access.
