School-based Drug Prevention
Challenges in Designing and Analyzing Social Experiments
Published in: Public Policy and Statistics: Case Studies From RAND (New York, N.Y.: Springer-Verlag, 2000), p. 9-28
Posted on RAND.org on December 31, 1999
Many important public-policy decisions rely on evaluating the effectiveness of social interventions. However, such evaluations pose substantial statistical challenges not faced by clinical trials in trying to reach valid conclusions. This chapter describes the design and analysis strategies adopted in a longitudinal experiment to evaluate a school-based drug-prevention program called Project ALERT. The experiment, conducted at 30 junior high schools in California and Oregon, assessed whether the curriculum reduced initiation and regular use of alcohol, cigarettes, and marijuana by adolescents. The chapter describes specific steps taken to address the following challenges: achieving full but realistic implementation of the curriculum; testing the program in diverse environments; obtaining well-matched experimental cells; obtaining reliable measures of use; adjusting for baseline differences among experimental groups; and accounting for intra-school correlation of outcomes.