NCJ Number
142327
Journal
Journal of Drug Issues Volume: 23 Issue: 2 Dated: (Spring 1993) Pages: 229- 250
Date Published
1993
Length
22 pages
Annotation
This examination of issues in the application of multiple-capture methods to drug-use prevalence estimation summarizes three published applications, discusses significant application issues, and compares three estimation approaches with the use of simulated data.
Abstract
The multiple-capture models are attractive because their statistical basis is clear and because they give specific numerical estimates with associated standard errors. The examples of applications described are the closed-population approach with data from two sources, the two-sample closed-population approach, and the three-sample closed-population approach. Application issues discussed include population definition, sampling data characteristics, model choice, and population dynamics. The three models compared are a set of six closed-population association models, the Jolly-Seber open-population model, and an open-population model based on a latent Markov process. Reasonable estimates were produced for the hypothetical population by all but the Jolly-Seber model. The Markov-based dynamic model was most accurate. Although it could not recover the exact frequencies, its treatment- corrected form came within 10 percent of the true values, and it correctly recovered the pattern of increases or decreases. The closed-population models also provided estimates of cumulative population size that were within approximately 15 percent of the true values. 3 tables and 20 references