Segmentation for path models and unobserved heterogeneity: The finite mixture partial least squares approach

Partial least squares-based path modeling with latent variables is a methodology that allows to estimate complex cause-effect relationships using empirical data. The assumption that the data is collected from a single homogeneous population is often unrealistic. Identification of different groups of consumers in connection with estimates in the inner path model constitutes a critical issue for applying the path modeling methodology to form effective marketing strategies. Sequential clustering strategies often fail to provide useful results for segment-specific partial least squares analyses. For that reason, the purpose of this paper is fourfold. First, it presents a finite mixture path modeling methodology for separating data based on the heterogeneity of estimates in the inner path model, as it is implemented in a software application for statistical computation. This new approach permits reliable identification of distinctive customer segments with their characteristic estimates for relationships of latent variables in the structural model. Second, it presents an application of the approach to two numerical examples, using experimental and empirical data, as a means of verifying the methodology's usefulness for multigroup path analyses in marketing research. Third, it analyses the advantages of finite mixture partial least squares to a sequential clustering strategy. Fourth, the initial application and critical review of the new segmentation technique for partial least squares path modeling allows us to unveil and discuss some of the technique's problematic aspects and to address significant areas of future research.