PLS Path Modeling represents a useful tool to model complex causal relations between blocks of variables in customers' preferences studies. When customers do have different behaviors, models accounting for unobserved heterogeneity permit to outline targeted and more efficient strategies, while a single model for all the customers might hide relevant differences in their behaviors. Unobserved heterogeneity implies to identify segments of units that exhibit similar behaviors, i.e. a response-based classification. This approach requires to look for latent segments as well as to estimate 'local' models. Different methods have been proposed for capturing unobserved heterogeneity in PLS Path Modeling. In this paper, the REsponse Based Units Segmentation (REBUS-PLS) approach is presented. REBUS-PLS is a distribution-free approach that aims to yield a classification of units which leads to 'local' models with an improved predictive performance as compared to a global model.
TRINCHERA, L., ROMANO, R. and ESPOSITO VINZI, V. (2007). Latent Segments Detection in PLS Path Modeling: A Tool to Capture Unobserved Heterogeneity in Customers' Preferences.