A probability sample of gay urban males: The use of two-phase adaptive sampling

If a target population, such as gay males, is relatively rare and nor easily identifiable it can be expensive and technically difficult to survey. Often, the population data available for planning such a survey is of suspect reliability. These obstacles often Lead to the abandonment of careful probability sampling relying instead on convenience samples such as organization membership lists or patrons of gay bars. The use of such haphazard sampling seriously undermines the credibility of surveys to estimate population characteristics or measure behaviors. This paper describes the implementation of a two-stage telephone survey design that while complex, achieves a probability sample of gay males in four major cities at reasonable cost. The design is based on an adaptive sampling approach which uses data from early stages of the survey to improve the efficiency of subsequent stages. Both the practical and theoretical issues involved in the design are discussed.