Analytica Chimica Acta 760 (2013) 34–45 Contents lists available at SciVerse ScienceDirect Analytica Chimica Acta jo u rn al hom epa ge: www.elsevier.com/locate/aca Predictive-property-ranked variable reduction in partial least squares modelling with final complexity adapted models: Comparison of properties for ranking  Jan P.M. Andries a , Yvan Vander Heyden b , Lutgarde M.C. Buydens c,∗ a Department of Life Sciences, Avans Hogeschool, University of Professional Education, P.O. Box 90116, 4800 RA Breda, The Netherlands b Department of Analytical Chemistry and Pharmaceutical Technology, Center for Pharmaceutical Research, Vrije Universiteit Brussel-VUB, Laarbeeklaan 103, B-1090 Brussels, Belgium c Radboud University Nijmegen, Institute for Molecules and Materials, Toernooiveld 1, 6525 ED Nijmegen, The Netherlands h i g h l i g h t s ◮ Variable reduction using the PPRVR- FCAM method is investigated. ◮ Performance of individual and com- bined predictor-variable properties is studied. ◮ Selective and predictive perform- ances of resulting models statisti- cally compared. ◮ Absolute PLS1 regression coefficient and its significance are most effec- tive. g r a p h i c a l a b s t r a c t Selected variables after variable reduction by the PPRVR-FCAM method, using individual and combined predictor-variable properties. a r t i c l e i n f o Article history: Received 7 August 2012 Received in revised form 31 October 2012 Accepted 8 November 2012 Available online 16 November 2012 Keywords: Variable reduction Partial least squares Predictor-variable properties Final complexity adapted models a b s t r a c t The calibration performance of partial least squares regression for one response (PLS1) can be improved by eliminating uninformative variables. Many variable-reduction methods are based on so-called predictor-variable properties or predictive properties, which are functions of various PLS-model param- eters, and which may change during the steps of the variable-reduction process. Recently, a new predictive-property-ranked variable reduction method with final complexity adapted models, denoted as PPRVR-FCAM or simply FCAM, was introduced. It is a backward variable elimination method applied on the predictive-property-ranked variables. The variable number is first reduced, with constant PLS1 model complexity A, until A variables remain, followed by a further decrease in PLS complexity, allowing the final selection of small numbers of variables. In this study for three data sets the utility and effectiveness of six individual and nine combined predictor-variable properties are investigated, when used in the FCAM method. The individual prop- erties include the absolute value of the PLS1 regression coefficient (REG), the significance of the PLS1 regression coefficient (SIG), the norm of the loading weight (NLW) vector, the variable importance in the projection (VIP), the selectivity ratio (SR), and the squared correlation coefficient of a predictor variable with the response y (COR). The selective and predictive performances of the models resulting from the use of these properties are statistically compared using the one-tailed Wilcoxon signed rank test.  Paper presented at the XIII Conference on Chemometrics in Analytical Chemistry (CAC 2012), Budapest, Hungary, 25–29 June 2012. ∗ Corresponding author. Tel.: +31 24 3653180; fax: +31 24 3652653. E-mail addresses: lbuydens@science.ru.nl, B.Loozen@science.ru.nl (L.M.C. Buydens). 0003-2670/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.aca.2012.11.012