Publications & Resources

Measurement Error in Multilevel Models of School and Classroom Environments: Implications for Reliability, Precision, and Prediction

May 2013

Jonathan Schweig

Measuring school and classroom environments has become central in a nationwide effort to develop comprehensive programs that measure teacher quality and teacher effectiveness. Formulating successful programs necessitates accurate and reliable methods for measuring these environmental variables. This paper uses a generalizability theory framework to compare and contrast four widely used approaches for accounting for measurement error in school and classroom level variables. Then, this paper uses two empirical examples to demonstrate how each of these approaches lead to different conclusions about measurement precision, and influences the conclusions about relationships between the environmental variables and policy-relevant outcomes. Additionally, this paper shows how one widely used model may misrepresent the structure of the data in many survey administration scenarios.

Schweig, J. (2013). Measurement error in multilevel models of school and classroom environments: Implications for reliability, precision, and prediction (CRESST Report 828). Los Angeles: University of California, Los Angeles, National Center for Research on Evaluation, Standards, and Student Testing (CRESST).