In an ideal world, school-based research would involve representative sampling of schools and students and randomization of individual students to treatment conditions. Here the units of assignment, observation, and analysis would be the same and the investigator would have confidence in the internal and external validity of the comparisons of interest. In the real world, convenience sampling and randomization of schools are the norm. Unfortunately, the former poses problems for external validity and the latter poses problems for internal validity; both pose problems for the analysis. As a result, the investigator must be both creative in analyzing the data and cautious in interpreting the results. After a brief summary of the validity problems, the authors will discuss the analyses reported in this issue from two different perspectives on how data from studies such as these might be analyzed.