Recent federal regulations requiring random drug testing of all transit agency employees are now being tested in the courts. However, many managements are still planning procedures for testing at least a portion of their work force for substance abuse. In devising appropriate sampling procedures, management must ensure that four distinct goals are met: fairness to all employees, unpredictability of who will be tested each day, maintenance of service, and economic efficiency. The goal of maintaining service may require a limitation on the number of employees sampled from any one location or occupational category in any one day. Simple random sampling, however, providesno guarantee of such a limit. A weighted random sampling technique is described. It was developed for the massachusetts bay transportation authority and allows constraints on the number of employees in various categories sampled yet maintains fairness in the sense that all employees subject to random drug testing have equal probability of being sampled. Whereas the proposed procedure is complicated compared with simple random sampling, it can be implemented and run on a personal computer. Implementation requires that a set of weights be computed for the work force and that the weights be used in the daily sampling procedure. The use of the weights on a daily basis is straightforward and requires only slightly more computational effort than construction of a simple random sample. Determination of the weights is more computation intensive but would typically be required only on a monthly or quarterly basis. This paper appears in transportation research record no. 1322, Large vehicle safety: transit and trucks 1991
Abstract