[Multilevel Analysis as a Tool to Analyze Research Questions in Rehabilitation Science]

Studies in rehabilitation science often investigate data representing different hierarchical data levels. Hierarchical sample structures prevail if single cases (e. g. rehabilitation patients) are grouped or are members of grouped entities or clusters (e. g. teams, groups, institutions), respectively. Analyzing data of rehabilitation patients treated in different institutions it has to be regarded that belonging to a specific institution (level 2; cluster-level) may be associated with individuals' (level 1) characteristics systematically. If rehabilitation patients take part in different training groups, specific characteristics of the training groups (level 2) may affect treatment effects of individuals (level 1) significantly. Multilevel analysis or hierarchical linear models allow for simultaneously modelling such multilevel data structures in an integrative and comprehensive manner. In this article it will be shown which specific aspects concerning hierarchical sampling procedures, data organization, specification of research hypotheses, statistical data analysis as well as interpretation of study results have be regarded to model potential clustering effects appropriately.