3rd IASC world conference on
Computational Statistics & Data Analysis
Amathus Beach Hotel, Limassol, Cyprus, 28-31 October, 2005
 
Title:  Partial Least Squares: A Framework for Data Analysis and Statistical Modeling

Description:

PLS (Partial Least Squares) Regression is a statistical modeling technique with data analysis features linking a block of response variables to a block of explanatory variables. This method is feasible even in critical situations often encountered in real practice when, for instance, there are missing data or too few observations or too many variables or a too strong correlation between the explanatory variables. The principle of the PLS algorithm may be similarly used in order to yield extensions of PLS regression to the case of logistic regression or, more generally, to PLS generalised linear regression. Yet, PLS regression is only one of the methods within the more general PLS framework. PLS regression may be actually meant as a specific case of the PLS approach to structural equation modeling when only two blocks of variables are available. This general approach studies a system of linear relationships between latent (non observable) variables. Each latent variable is described by a set of manifest (observable) indicators. The nature of this approach is rather exploratory and data-driven than confirmatory. It is usually referred to as a "soft" modeling approach as no normality hypothesis is required, there is no constraint on the measurement scale of the manifest indicators and the number of observations may be limited with respect to the number of variables. The features of PLS-based methods make them very interesting for further methodological developments and for applications in several domains. The PLS track will provide an expository review of the methodologies comprised in the PLS framework with a specific focus on the critical assessment of the most recent developments with new statistical and computational topics, the presentation of fresh ideas on the PLS theory and methodology, the comparison of these methods with other competitors or alternatives, and the estimation of the future directions of research in the field.

Focus:

PLS Regression and related extensions
PLS Path Modeling and Alternative Approaches
PLS vs. Methods for Multiblock Analysis
Missing Data treatment in PLS
Classification issues in PLS
New Validation and Interpretation tools for PLS
Outlier Identification in PLS
PLS for Stochastic Processes
PLS with very large datasets
PLS for complex (L, U, Domino, Networks) data structures

Co-Chairs:

Carlo Lauro
Dipartimento di Matematica e Statistica
Facolta di Economia
Universita degli Studi di Napoli "Federico II"
Via Cintia, 26 - Complesso Monte SantAngelo
80126 Napoli - Italy
E-mail: carlo.lauro@unina.it

Vincenzo Esposito Vinzi
Dipartimento di Matematica e Statistica
Facolta di Economia
Universita degli Studi di Napoli "Federico II"
Via Cintia, 26 - Complesso Monte SantAngelo
80126 Napoli - Italy
E-mail: vincenzo.espositovinzi@unina.it
Wynne W. Chin
Bauer Faculty Fellow
Department of Decision and Information Sciences
C.T. Bauer College of Business
334 Melcher Hall, room 280D
University of Houston
Houston, Texas 77204-6282 USA
E-mail: wchin@uh.edu
Aloke Phatak
CSIRO Mathematical and Information Sciences
Private Bag No. 5
Wembley, WA 6913
Australia
E-mail: Aloke.Phatak@csiro.au