UCLA

Performance-targeted interventions, typically based on student test performance, arean important tool in improving educational outcomes. These types of interventions are often applied at the school level, where low-performing schools are selected for participation. However, typical school effects methods for understanding school performance do not directly identify the low-performing schools that would benefit the most from additional support. Additionally, typical school effects methods do not differentiate school performance based on important aspects of the curriculum. This dissertation fills this gap in school effects methods by proposing the Multilevel Diagnostic Item Response (MD-IR) model. The MD-IR model is a multilevel, confirmatory mixture item response theory model that incorporates strategic constraints in order to differentiate schools, and students within schools, based on the aspects of the curriculum that would be most relevant for a performance-targeted intervention. By incorporating latent classes, the MD-IR model classifies schools as high- or low-performing, and as such, identifies schools most in need of support. The formulation of the MD-IR model is presented, along with a detailed empirical example demonstrating its application in the context of international educational development using data from PISA for Development. Results from the empirical example illustrate the utility of this model and its promise in filling this identification gap in the school effects literature.