Partial Least Squares Is to Lisrel as Principal Components Analysis Is to Common Factor Analysis

The decision of whether to use PLS instead of a covariance based structural equation modeling technique such as LISREL for causal modeling can be assisted by looking at the differences between principal components analysis and common factor analysis. Through such a process, this paper outlines the need for PLS users to shift from merely estimating model parameters to that of including measures of predictive relevance. Unless the communality is high and the indicators per construct are large, the PLS parameter estimates for construct loadings will likely have a homogenization and overestimation bias. Conversely, the structural paths tend to be underestimated. Commentator's Biography Wynne W. Chin is Associate Professor in Management Information Systems at the University of Calgary. He received an AB in biophysics from U. C. Berkeley, an MS in biomedical engineering from Northwestern University, and an MBA and Ph.D. in computers and information systems from The University of Michigan. He is currently completing the development of PLS-Graph a Windows based software for performing Partial Least Squares analysis. His research interest involves the impact of individual attitudes and social networks on user acceptance of new information technology. His other research interests include methodological issues in information systems research, group support systems, and end user computing. Mailing Address: Faculty of Management, University of Calgary, Calgary, Alberta T2N 1N4, Canada; Phone: (403) 220-3732; Fax (403) 282-0095; E-mail: [email protected] on Internet. This is a copy of my original submission that was published as Chin, W. W. (1995). Partial Least Squares is to LISREL as principal components analysis is to common factor analysis. Technology Studies, 2, pp. 315-319. The final publication was less complete than this document since it omitted the abstract, made several mathematical typos, and left out two references. Otherwise, everything else remains the same.