Hierarchical linear models - their advantages and reasons for application

Independence of observations is one of the key assumptions underlying regression analysis and other methods based on the general linear model. The assumption of independence of observations is met when the result of the outcome variable obtained by a person in a data set is not dependent on results of other persons. This article introduces the hierarchical linear modeling (HLM) - the statistical method that is recommended when there is a real chance that the assumption of the observations᾿ independence is violated. The structure of the article is threefold. In the first part the authors present basic methodological reasons for applying HLM method, stressing its advantages in comparison to the traditional regression analysis based on the ordinary least squares estimation. The second part introduces the most important theoretical notions underlying the hierarchical model - the division into fixed and random effects, the multilevel data structure (including cross-level interaction), and the specific approach to variance components. The third part of the article shows two empirical examples of application of HLM, including a detailed interpretation of their results. peerReviewed publishedVersion