The Use of Archival Data To Select and Assign Schools in a Drug Prevention Trial

School-based prevention trials typically face large variations in school composition and levels ofpreintervention behavior. Such variations may inhibit efforts to maximize internal and external validity. This article presents a method for using school-level attributes in sample description and random assignment of schools to conditions. The utility of the archival data was examined using multivanate canonical techniques. The authors found that a small set of attributes could efficiently predict the observed school-level variations in smoking and other drug use. It is recommended that these attributes be used by other school-based studies to enhance experimen talfield studies.