Using multilevel regression to examine hierarchical data: investigating differences in reading performance between immigrant and native-born children / El uso de la regresión multinivel en el examen de datos jerárquicos: un estudio sobre las diferencias existentes en el rendimiento en la lectura entre niños de origen inmigrante y niños nativos

Abstract

A persistent problem in examining culture and education is the entanglement of individual and group aspects of norms, beliefs and behaviour. The challenge often concerns the proper specification and testing of theories about how such group and individual processes work with methods that can address the complexity of hierarchical data sets. Multilevel modelling is an attractive analytic approach for studying individual and group processes in educational settings because it facilitates the incorporation of substantive theory about how such processes work within the clustered sampling schemes typical of educational data. In this article, we illustrate how to incorporate two-level multilevel regression modelling into an example investigation of the reading achievement of immigrant students who receive English language services versus their native-born peers who do not receive these services. We conclude with a few ways in which multilevel modelling can enhance the investigation of multilevel theories regarding how cultural and organizational processes work in educational research.