Partial Least Squares Path Modeling (PLS-PM) is a statistical approach, with an increasing popularity in several areas, for modeling complex multivariable relationships among observed and latent variables. From the standpoint of structural equation modeling, we look at PLS-PM as a component-based approach where the concept of causality is formulated in terms of linear conditional expectations, thus privileging a prediction-oriented discovery process to the statistical testing of causal hypotheses. From the standpoint of data analysis, we mean PLS-PM as a flexible approach to the analysis of multiple blocks of variables that are available for the same set of samples. This approach shows how the ‘data-driven’ tradition of multiple-table analysis can be merged with the ‘theory-driven’ tradition of structural equation modeling so as to allow researchers to run the analysis in light of the current knowledge on the conceptual relationships between tables and with known optimizing criteria. In both frameworks, PLS-PM may enhance potentialities even further, and provide effective added value, when exploited in the case of formative relationships between manifest variables and their respective latent variables. In such a case, we show how PLS regression and PLS-PM can profitably interplay and coherently merge, thus permitting further developments with applications to marketing.
ESPOSITO VINZI, V. et FAHMY, T. (2007). Recent Developments in PLS Path Modeling: Methodological Issues and Applications.