In field research, randomized studies are relatively uncommon in criminological research as they encompass a range of ethical, legal and practical difficulties. Randomized control trials (RCTs) are the optimum choice when the integrity of the intervention can be precisely controlled. Analysis based on initial allocation to condition within an RCT, regardless of whether the treatment was actually delivered, is called intention to treat (ITT). Alternatively, analysis based on treatment received (TR), considers outcomes when treatment is completed. Quasi-experimental designs offer an alternative to an experimental design by maintaining internal validity through methodological or statistical control of differences between experimental and control conditions. Weisburd, Lum and Petrosino compared outcomes from fully randomized studies and other experimental designs in crime prevention studies. Weisburd and his colleagues noted an inverse relationship between research design and outcome, suggesting that non-randomized designs may introduce a bias in favour of treatment.