Possibilistic PLS path modeling: A new approach to the multigroup comparison

Abstract

Structural equation models are reference techniques for measuring cause-effect relationships in complex systems. In many real cases observations are a priori grouped into homogeneous segments according to a specific characteristic, so that different models can be assessed for each segment. The present paper proposes to adopt an Euclidean metric based on the model parameters: the aim is to determine differences among models. However, estimated models assess the relation structures in different proportions, i.e. the residual component can vary with respect to the different models. In order to overcome this shortcoming, the present work proposes alternative models with fuzzy parameters.