Hierarchical linear models distinguish between the individual and the group levels of data. Hence, they are often referred to as multilevel models. It is easiest to think of hierarchical linear models as special regression models that allow simultaneous investigation of the respective roles that individual and group characteristics play in the attainment of treatment goals. A hierarchical linear model can be fit to the data to investigate the contribution of individual characteristics to young people's end of treatment attainment and to determine how these relations are influenced by group characteristics. To simplify, people think of data as consisting of only two levels: individuals nested within treatment groups; and groups. The data are thus described by two models: the first model is based on individual data. It describes the prediction of individual outcome from individual characteristics and is referred to as the within-unity, or unit level mode. The second model expresses the variability in regression coefficients as a function of group-level variables. This model is referred to as the group-level, or between-unit mode. The majority of research studies on group therapy have ignored the group effect. Given the fact group characteristics are the prime reason for selecting this treatment modality, it appears important to include these characteristics in any model for treatment effectiveness. (ABL)