A Methodological Comparison between PLS Path Modeling and Generalized Structured Component Analysis

PLS Path Modeling and Generalized Structured Component Analysis are two component-based approaches to Structural Equation Models. Both methods aim at estimating the causal relationships linking two or more latent variables by means of a set of observed indicators. Moreover, both methods define the latent variable as a linear combination of its own observed variables, i.e. as a component. The main difference between the two approaches is in the estimation of the model parameters. In fact, PLS Path Modeling uses an iterative algorithm based on a series of interdependent regressions, while the Generalized Structured Component Analysis defines a unique algebraic formulation for the model and uses an Alternative Least Square algorithm. Despite several studies have been performed to compare the PLS Path Modeling and the covariance-based approaches to Structural Equation Modeling, only a recent simulation study involves also the Generalized Structured Component Analysis (Hwang et al., 2010). Here, a new simulation study is presented in order to asses the performances of both PLS Path Modeling and Generalized Structured Component Analysis. We discuss the links between Generalized Structured Component Analysis and the Maximum Sum of Explained Variance method (Glang, 1988), as well as the links between the Glangs’s method and the PLS Path Modeling. Moreover, in Generalized Structured Component Analysis the measurement model seems to play a major role when estimating the model parameters.