Hierarchical Linear Modeling

Hierarchical linear modeling is also known as using multilevel models, variance component models, or random effect models. These models are used when data have a hierarchical or clustered structure. Hierarchical structures are the norm in the social sciences; for example, patients are treated within hospitals, people live in [Page 774]households, employees work within companies, and children learn within the same classrooms. This structure introduces dependence into the data, as units observed within clusters are more similar than units chosen at random from the population.

Traditional multiple regression techniques assume that observations are independent. Ignoring the clustered structure of the data leads to an underestimation of the standard errors of regression coefficients, leading to an overstatement of statistical significance. However, there are other methods for adjusting standard ...