Skip to Main content Skip to Navigation

Contributions aux modèles d'équations structurelles à variables latentes

Abstract : Path models with latent variables are complex statistical models able to interpret interactions between unobservable concepts. The two leading estimation methods, covariance structure analysis (LISREL) and partial least squares path modeling (PLS-PM), appear complementary rather than competitive. In this work, we first introduce models and methods related to structural equations modeling and, then, we answer several questions on path models with latent variables. We present algorithms to construct the initial measurement and structural models. We introduce methods based on permutation tests for multi-group comparison. Approaches based on variables transformation to find nonlinear relationships are exposed. Finally, methods to handle specific cases of missing data patterns are presented. We first focus on existing theories and, then, introduce new approaches. Finally, the new methods are applied within the framework of customer satisfaction and loyalty to the French electricity sector.
Complete list of metadatas

Cited literature [268 references]  Display  Hide  Download
Contributor : Emmanuel Jakobowicz <>
Submitted on : Wednesday, March 26, 2008 - 3:14:39 PM
Last modification on : Wednesday, March 20, 2019 - 4:52:07 PM
Long-term archiving on: : Tuesday, September 21, 2010 - 3:51:56 PM


  • HAL Id : tel-00207990, version 2



Emmanuel Jakobowicz. Contributions aux modèles d'équations structurelles à variables latentes. Mathématiques [math]. Conservatoire national des arts et metiers - CNAM, 2007. Français. ⟨tel-00207990v2⟩



Record views


Files downloads