A continuum regression approach based on two parameters, namely the number of latent variables to be introduced in the model and a tuning parameter is proposed. A noteworthy feature of this approach is that it is straightforward and simple. It brings several procedures under the same umbrella. These procedures encompass Ordinary least squares (OLS) regression, Partial least squares regression (PCR) and Ridge regression. Interesting properties related to the method are also discussed. This makes it possible to highlight the rationale behind the method of analysis by showing that it aims at realizing a compromise between achieving a good fit and setting up a stable model. Copyright © 2006 John Wiley & Sons, Ltd.
[1]
S. Wold,et al.
The multivariate calibration problem in chemistry solved by the PLS method
,
1983
.
[2]
B. M. Wise,et al.
Canonical partial least squares and continuum power regression
,
2001
.
[3]
J. Friedman,et al.
A Statistical View of Some Chemometrics Regression Tools
,
1993
.
[4]
Petre Stoica,et al.
Partial Least Squares: A First‐order Analysis
,
1998
.
[5]
Henk A. L. Kiers,et al.
Principal covariates regression: Part I. Theory
,
1992
.
[6]
David A. Belsley,et al.
Conditioning Diagnostics: Collinearity and Weak Data in Regression
,
1991
.