State-of-art on PLS Path Modeling through the available software.pdf

State-of-art on PLS Path Modeling
through the available software
Prepared by
Yves Marie Chatelin, IdE France
Vincenzo Esposito Vinzi, DMS Italy
Michel Tenenhaus, HEC France
June 2002
State-of-art on PLS Path Modeling
Table of Contents
I. The ECSI model 3
II. The PLS Path model 5
1. Relation between the manifest variables and the latent variables (outer model) 5
2. Normalization of the latent variables 6
3. Relation between the latent variables (inner model) 6
III. Partial Least Squares (PLS) Estimation of the ECSI model 7
1. Manifest variables standardization 7
2. Latent variables estimation 8
. External estimation Yj of the standardized latent variable (j – mj) 8
. Internal estimation Zj of the standardized latent variable (j – mj) 8
. Estimation of the weights wjh’s 9
. Estimation of structural equations 10
IV. Model Validation 10
1. Blindfolding approach in LVPLS 10
2. Resampling in PLS-Graph 12
V. Missing Data Treatment 14
VI. Use of LVPLS and PLS-Graph 15
The case of the ECSI model (with missing data) for a mobile phone
provider
1. Options selection 15
2. The ECSI model for a Mobile Phone Provider 15
3. The results of Resampling Procedures on Customer Satisfaction 17
4. The results of Resampling Procedures on Path Coefficients of the model 19
REFERENCES 19
ANNEX 1 21
The program code for the Lohmöller LVPLS program
(when some data are missing)
ANNEX 2 22
The results of the Lohmöller LVPLS program
(ECSI model with missing data)
The results with Blindfolding on Customer Satisfaction 23
(Redundancy munality criterion, 30 groups)
Model Validation 25
ANNEX 3 26
Resampling in PLS-Graph (ECSI Model : Latent Variable Customer Satisfaction)
ANNEX 4 29
Glossary
2
State-of-art on PLS Path Modeling
State-of-art on PLS Path Modeling through the available software
Yves-Marie Chatelin, Vincenzo Esposito Vinzi & Michel Tenenhaus
Abstract
The purpose of this paper is to present PLS