Two Multivariate Methods for Analyzing Small Group Interaction

Abstract

This article compares the advantages and disadvantages of two multivariate methods for the analysis of small grolup communication data, log-linear contingency table analysis and multivariate analysis of variance, in the context of ani experimental design. The independent conditions were discussion procedure, task solution multiplicity, and discussion topic. Participant communication behaviors, the dependent variables, were coded into a mix of 15 task social, and procedural categories. The analyses of communication indicated that log-linzear analyasis produced thle most complex models for various behaviors, including both single-factor and multifactor termns. Multivariate analysis of variance (MANOVA) findings at tlte grolup level provided the most parsimonious explanations, with a main effect of discussion procedure. Some MANOVA results at the individual level were similar to loglitnear mnodels in their complexity, whereas others duplicated the results of the group-level MANOVA. The differing outcomes from these methods point to several design considerations that should be taken account when choosing a statistical test for group communication data.