|
Activity Number:
|
471
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #301784 |
|
Title:
|
The PLS-CBSEM Debate and an Investigation of Alternative Modeling Methods
|
|
Author(s):
|
Irene R.R. Lu*+ and D. Roland Thomas and Ernest Kwan and Marzena Cedzynski
|
|
Companies:
|
York University and Carleton University and York University and Carleton University
|
|
Address:
|
4700 Keele Street, Toronto, ON, M1V 4J1, Canada
|
|
Keywords:
|
Partial Least Squares ; Structural Equation Modeling ; Latent Variable Modeling ; Path Modeling ; PLS ; SEM
|
|
Abstract:
|
Partial least squares (PLS) is sometimes used as an alternative to covariance-based structural equation modeling (CBSEM). We identified and examined advantages of PLS over CBSEM commonly cited by PLS users from a broad range of social and behavioral disciplines. We further compared the performance of PLS and CBSEM to other latent variable modeling methods through a series of Monte Carlo simulations. The performance criteria comprised parameter estimation bias, standard error of parameter estimates, confidence interval coverage, and statistical power. Based on the simulation results we provide some recommendations on choosing methods for latent variable modeling.
|
