PLS Path Modeling is a component-based technique to study linear relations among multiple blocks of indicators in a structural equation model. This approach assumes homogeneity among statistical units thus preventing the detection of latent classes. This paper proposes to combine the fuzzy approach to PLS Path Modeling in order to overcome a few shortcomings of already existing methods, such as the strong normality assumption on latent variables in the case of finite mixture models, or the lack of an optimizing criterion in the Typological approach. As a matter of fact, the fuzzy approach embeds residual information inside the model coefficients (expressed as intervals) thus permitting to search for latent classes by referring to the fuzziness of coefficients.
ESPOSITO VINZI, V., TRINCHERA, L. and ROMANO, R. (2007). Fuzzy PLS Path Modeling for Latent Class Detection.