This chapter examines how a particular strategy for analyzing evaluation data, intent-to-treat (ITT) analyses, may underestimate the true effects of interventions. Such underestimation of intervention effects can profoundly influence policies for prevention and treatment of children's mental health problems, which can in turn lead to negative consequences for children's healthy development. However, evaluating treatment is a complicated issue because poorer outcomes for some may be due to characteristics of the participants, such as low motivation or chaotic family conditions, rather than qualities of the intervention. ITT analyses purposely ignore these nonrandom sources of variance. Using ITT analyses, evaluations of programs to reduce oppositional defiant disorders and conduct disorders in children and adolescents have consistently revealed that cognitive-behavioral interventions have the most promise and clearest evidence for efficacy, with effect sizes on outcome analyses in the moderate to large range (from 0.3 to over 1.0). These interventions usually involve behavioral parent training but also can include social problem-solving skills training, anger management training, and social skills training with the children. Using analyses of the effects of an intervention designed to reduce children's externalizing behavior problems and thus their risk for later delinquency and substance use as an example, the authors compare how propensity analyses and three types of complier average causal effect (CACE) analyses fare in comparison to traditional ITT analyses and often-used as-treated analyses. Although these techniques have been presented as an alternative and possible improvement, analyses of compliance have not considered the consequences of how the criteria for compliance are determined for a particular intervention, nor do they account for whether the compliance analyses produce different effects at different levels of compliance. Therefore, the present example also compares two levels of compliance, one representing a criterion of at least minimal compliance with the intervention and a second representing a high level of attendance and compliance. The general conclusion as a result of comparing these multiple strategies is that how one specifies parents' compliance in an evaluation design for a preventive intervention affects the interpretation of findings of program efficacy. The authors suggest that future research should aggressively pursue methods for setting optimal thresholds in analytic approaches that extend beyond ITT. (Contains 1 table.)