ABSTRACT
Previous chapters in this text have addressed the use of multivariate analysis of variance (MANOVA) and hierarchical linear modeling or, more generally, multilevel modeling. Traditional applications of these procedures have limitations that restrict their use. In particular, standard use of MANOVA assumes that responses of individuals are independently distributed, an assumption that may be violated when participants are nested in organizations or settings (such as students nested in schools, clients nested in therapists, workers nested in workplaces). When such dependence is present, use of MANOVA may result in unacceptably high type I error rates associated with the effects of explanatory variables, as detailed in Chapter 6. For its part, multilevel modeling accommodates the dependence arising from such clustered data that MANOVA does not. However, standard multilevel modeling is able to incorporate only one dependent variable from units, often participants, at the lower level. Thus, such use of multilevel modeling is not able to take advantage of the benefits associated with multivariate analysis that have been described previously in this book.
