Hierarchical Linear Models
Multilevel analysis is a general statistical approach for analyzing data that vary at more than one level. Hierarchical linear models are tools of multilevel analysis which enable researchers to analyze data that have a “hierarchical” structure. In basic terms, hierarchy refers to situations in which the individual subjects of a study can be categorized or ordered into distinct levels (or hierarchies) which share certain traits that may influence the study outcome. Hierarchical linear modeling allows researchers to account for such shared traits or variance in their data.
Hierarchical linear models are now frequently used to analyze educational datasets, which often have hierarchical patterns of variability. These patterns emerge because data for the educational sector may be organized at the levels of students, classrooms, grades, schools, and school districts. For example, educational datasets frequently encompass information from students from several schools in a district but the traits of data from students attending the same school will tend to be correlated. Within data from the same school there may in turn be correlations, for example, at the grade level and at classroom level. Hierarchical linear modeling allows for the analysis of relationships between data at, for example, the student level, while also accounting for variability at the classroom and school levels. Hierarchical linear modeling is not a particularly new methodology. However, recent improvements in computing power and software allow researchers to make greater use of this complex form of statistical analysis.