Multilevel Modeling for Research in Group Work

Abstract

Researchers in group counseling often encounter complex data from individual clients who are members of a group. Clients in the same group may be more similar than clients from different groups and this can lead to violations of statistical assumptions. The complexity of the data also means that predictors and outcomes can be measured at both the client and the group level. Researcher questions may focus on variables at the client level or the group level, or the interaction of client and group level variables. In this article, we introduce multilevel modeling as a tool that can be used both to account for the complex structure of the data and to incorporate variables at both the client and group levels. A published group counseling study is used as an example.

Keywords:

• group outcomes
• group work
• multilevel modeling
• nested data structures
• statistical models for group counseling