Definition of Multilevel (hierarchical linear) Modeling

A form of linear regression (also known as hierarchical linear modeling [HLM]). The defining feature of multilevel modeling is the allowance for the variance of the outcome variable to be analyzed at multiple levels, whereas simple and multiple linear regression models all have variance at a single level. The data that bests suit multilevel modeling are data that are naturally nested (e.g., students nested within schools). There is some variance that is accounted for between schools, in addition to the variance accounted for between the students nested within those schools. Hierarchical modeling allows examination of those nat urally occurring levels of variance and the calculation of the proportion of variance that is within each level of the data (intraclass correlation).