Partial Least Squares Path Modeling

Abstract

Partial Least Squares Path Modeling (PLS-PM) is one of the PLS techniques with a great modeling power. It is considered as a multivariate technique of second generation, thus providing an insight scheme by combining causal modeling with data analysis features. PLS-PM is a statistical method that has been developed for the analysis of structural equation models with latent variables, specially designed to provide an alternative approach to the most well-known LISREL models. As opposed to the covariance-based approach, PLS is prediction oriented aiming to obtain estimates of latent variables for prediction purposes. The theoretical foundations of PLS are strongly related to the Principal Components framework, especially with the algorithms used to solve principal components based problems. In the PLS field we find the Nonlinear Iterative Partial Least Squares (NIPALS) algorithm developed to calculate principal components. For this reason, PLS- PM can be viewed as a component-based approach to structural equation modeling in the sense that the goal is to estimate the latent variables as components of the manifest variables. This chapter begins with an overall review of the PLS conceptual backgrounds, particularly a brief description of the Principal Components Analysis followed by the explanation of the NIPALS algorithm. Then, the PLS-PM method is presented in a detailed way. In the last section we briefly discuss the Generalized Structure Component Analysis (GSCA) which has been recently proposed as an alternative to PLS-PM in the component-based approaches to SEM.