WRTPLS, Selection of the Number of Components in PLS - Partial Least Squares

Weight randomization test for the selection of the number of components in PLS models

Journal of Chemometrics. 2017;31:e2887.https://doi.org/10.1002/cem.2887
Thanh Tran, Ewa Szymańska, Jan Gerretzen, Lutgarde Buydens, Nelson Lee Afanador, Lionel Blanchet, "Weight randomization test for the selection of the number of components in PLS models", J. Chemometrics, Volume 31, Issue 5, May 2017
The Weight Randomization Test (WRT) is introduced for selection of the optimal number of components in PLS. The WRT aims to minimize the amount of irrelevant data entering a PLS model, using a newly defined model quality metric: the number of Consecutive Significant Components. The latter is determined using randomization tests, completely in accordance with the underlying theory of the PLS algorithm. This development is an important step toward fully automating PLS modeling. Two applications show the benefits of WRT.
Keywords: PLS, Regression, Selection of PLS Components, Model Quality

EXAMPLE:

% Example:
% load spectra
% x = NIR;
% y = octane;
%
% Get default options
% options = wrtpls('options');
% Run WRTPLS with default options
%[reg,xlds,ylds,wts,xscrs,ncomp] = wrtpls(x,y,options);
%
% Customized options
% options = wrtpls('options');
% options.figON = 1;
% options.numnonsigs = 1;
% options.maxcomp = 25;
% [reg,xlds,ylds,wts,xscrs,ncomp] = wrtpls(x,y,options);

In case of publication of any application of this method,please, cite the original work:
Thanh Tran, Ewa Szymańska, Jan Gerretzen, Lutgarde Buydens, Nelson Lee Afanador, Lionel Blanchet, Weight randomization test for the selection of the number of components in PLS models, Journal of Chemometrics. 2017;e2887.
DOI: https://doi.org/10.1002/cem.2887