Consistency Between Theoretical Interests in Collaborative Learning Studies and Methods of Statistical Analysis : A Review of Statistics for Hierarchical Data(<Special Issue>Collaborative Learning and Networked Communities)

Abstract

Statistical methods are used in studies of collaborative learning to analyze questionnaire and achievement test data. In collaborative group learning studies, the theoretical assumptions of collaborative learning, however, conflicts with the assumption made in the tests, namely that the samples are independent. In this paper, statistical methods for analyzing hierarchical data and clustered samples, which are consistent with the theoretical assumption of collaborative learning, are identified. Such methods are particularly suitable for analyzing aggregated data, and include regression analysis with robust standard error, hierarchical linear model (HLM), and multilevel covariance structural analysis (MCA). The special features of these methods are discussed.