When Total Randomization Is Impossible

Abstract

This article describes the design of the evaluation of Healthy for Life (HFL), an adolescent health promotion project involving students in 21 middle schools in Wisconsin. The original design was a blocked random assignment of 21 schools to one of three conditions. However, most of the interested schools could not accommodate the random design. A two-step alternative procedure allowed schools to select one of two treatment conditions, with random assignment to the control condition from either treatment condition. This randomized control group design nested within two self-selected treatment options is a viable alternative to total randomization.