Hierarchical Linear Models for the Analysis of Longitudinal Data with Applications from Hiv/aids Program Evaluation

In this paper, two examples of multilevel modeling as part of the analysis of data from HIV evaluation studies are presented. Strategies for teaching multilevel models for each type of data are discussed. The first, a panel study, uses multiple linear regression models to show how a hierarchical linear model can be developed. The second, a repeated cross-sectional design, uses simple analysis of variance models to show how a random coefficients model can be fit to the data. Complex multilevel models may be easier to understand and apply when broken down into these more familiar strategies. Analyses are presented using the HLM program and SAS.