The selection of test sites or a judgmental sample of test items is often influenced by political and administrative considerations. Yet the researcher would like to make this selection as objective as possible, and at the same time be guaranteed that the selections made are representative of the domain from which the test items are selected. Often the selection process takes into account demographic and economic statistics about an area gathered in a census or estimated from survey data. One way of approaching this problem is to use statistical distance measures. The measures described in this paper were used to calculate a distance for each site in a set under consideration from a standard specified by the researcher. However, when two or more test sites are being selected, this method may not yield the best combination of sites or items in regard to the distance of the combination from the standard. This paper examines conditions under which consideration of individual distances may be inappropriate. However, it is shown how the individual distances may still be used, while improving the selection process for combinations of test sites.
-Cowan, Charles D., and Spoeri, Randall K.,