A Study on Methods for Capturing and Treating Heterogeneity in PLS Path Models by Unit Segm.pdfVIP

A Study on Methods for Capturing and Treating Heterogeneity in PLS Path Models by Unit Segmentation: FIMIX-PLS, PLS-TPM and REBUS-PLS Christian M. Ringle University of Hamburg; School of Business, Economics and Social Sciences; Von-Melle-Park 5; 20146 Hamburg; Germany; e-mail: cringle@econ.uni-hamburg.de Silvia Squillacciotti Electricité de France; Research Development; 1 Avenue du Général de Gaulle; 92141 Clamart; France; e-mail: silvia.squillacciotti@edf.fr Laura Trinchera University of Naples; Department of Mathematics and Statistics; Via Cintia 27; Complesso Monte Sant’Angelo; 80126 Naples; Italy; e-mail: ltrinche@unina.it Introduction: PLS Path Model Based Segmentation Analyses in a PLS Path Modeling (PLS-PM) framework [TVCL05] usually do not address the problem of heterogeneity. The failure to account for heterogeneity may lead to ambiguous PLS results and, thus, to conclusions that are incomplete and ineffective. A first approach is to test if data on the manifest variables level belongs to a finite number of clusters by sequential segmentation strategies such as K- means or tree clustering. These methods, however, are unable to account for heterogeneity in the relationships between latent variables and are improper for forming groups of data with distinctive PLS path model estimates [JJD97]. Alternatively, theoretical knowledge allows to analyze the effect of moderating factors [HF07]. This kind of approach only accounts for heterogeneity of a limited number of relationships in the inner model and is restricted to the availability of moderating variables. As a consequence of these limitations in existing methodologie