Comparing PLS to Regression and LISREL: A Response to Marcoulides, Chin, and Saunders

We note that in cases where the model errors are not explicitly taken into account for the estimation of endogenous latent variables, a new approach proposed by Vittadini et al. (2007) would need to be used to appropriately determine the PLS model estimates. This is because in PLS the reflective scheme assumed for the endogenous latent variables is inverted (i.e., the model errors are not taken into account). Savalei and Bentler (2007) have also indicated that with this type of model, PLS estimates (for example, those obtained by regression on factor scores) are generally biased as estimators of regression among latent variables (p. 172).