It is a common practice in educational research to model treatment variables at the classroom level rather than at the student level. Students are naturally nested within classrooms in the educational system; therefore, randomly assigning students at the student level into treatment and control groups is often not practical. Data collected in educational settings are usually hierarchical in nature, for example, students nested within classrooms and classrooms nested within schools. Failure to consider the hierarchical structure of educational data could cause unreliable estimation of the effectiveness of school context or teacher quality on student learning outcomes and could misdirect educational policies and practices (Raudenbush and Bryk, 2002). Three major concerns should be taken into consideration when analyzing multi-level data: (a) unit of analysis, (b) statistical procedures, and (c) conceptual fallacy.