PLS Path Modelling (PLS-PM) is generally meant as a component-based approach to structural equation modelling that privileges a prediction oriented discovery process to the statistical testing of causal hypotheses. In case of formative relationships between manifest and latent variables, PLS-PM implies multiple OLS regressions. They might yield unstable results in case of strong correlations between manifest variables while being not feasible when the number of observations is smaller than the number of variables nor in case of missing data. We explore PLS regression (PLS-R) as an external estimation mode to overcome the mentioned problems while preserving formative relationships and being coherent with the component-based and prediction-oriented nature of PLS-PM. PLS-R is also fruitfully extended to the path weighting internal estimation scheme and the estimation of path coefficients.
ESPOSITO VINZI, V. (2008). A Partial Least Squares Comprehensive Environment: In: 7th International Conference on Social Science Methodology, RC33 - Logic and Methodology in Sociology.
Keywords : #Multicollinearity