ABSTRACT

After first reviewing basic concepts of the single-level multiple regression model, the chapter develops the necessary steps of conducting a multilevel regression analysis using IBM SPSS Mixed. The intent is to develop the rationale behind the specification of this general class of models in a relatively nontechnical manner and to illustrate its use in an applied research situation. The methods presented in this chapter provide a basis for applying multilevel-modeling techniques to a broader set of research problems. The chapter first contrasts predictive (i.e., the efficiency of the prediction and the parsimony of variables included in the prediction equation) versus explanatory (the correct specification and testing of a theoretical model that is under consideration) approaches to model building. It then presents several advantages of multilevel analysis over traditional single-level univariate and multivariate approaches. Finally, it develops a series of multilevel models involved in investigating randomly varying outcome parameters using an extended example. These typically include variation in the levels of the outcome (intercepts) and the strength of within-group relationships indicated by regression coefficients (slopes) across groups. Once researchers identify that variation exists in the parameters of interest, they can build models to explain this variation.