Empirical Likelihood for Combining Judgment and Probability Sample Data Across Space and Time
by
Stephen L. Rathburn
Department of Statistics, University of Georgia, Athens, Georgia 30602, USA

Programs that monitor aquatic resources typically employ judgment sampling designs in which sample sites are selected according to a number of often vaguely defined criteria. Such designs often yield biased estimates of population parameters, such as the total mass of a contaminant in the bottom sediment of lake waters. Only through the implementation of probability-based sampling designs can design-unbiased estimators be obtained. However, managers of water quality monitoring programs are reluctant to implement probability- based sampling designs out of fear of losing information from the historical data base. I shall consider an approach to combining historical judgment sample data with new probability sample data to predict what would have been obtained had a probability-based sampling design been employed from the beginning of the monitoring program. This approach requires an interval of overlap during which both judgment and probability sites are sampled to obtain maximum empirical likelihood estimators of weights for the judgment sample sites. Then the past population means can be estimated using a weighted average of the past judgment sample data. The properties of the resulting estimator are explored under realizations of a number of alternative spatio-temporal models.

Date received: October 9, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cafr-19.