A Comparison of First and Second Generation Multivariate Analyses: Canonical Correlation Analysis and Structural Equation Modeling

Abstract

This study illustrates that Structural Equation Modeling (SEM) provides a more accurate representation of the latent variables as assessed by the structure coefficients than Canonical Correlation Analysis (CCA). A heuristic data set of five independent variables representing the Piagetian tasks measured by The Inventory of Piaget’s Developmental Tasks (IPDT) and three dependent variables assessed by the Alabama Basic Competency Tests (BCT) were analyzed with both statistical methods. Although the relationship between IPDT and BCT was invariant, the corresponding latent variable structures differed for the two methods.