Dina Research School

Multivariate regression modeling using Partial Least Squares

Frans van den Berg

Another important tool in chemometrics is Partial Least Squares regression for model building between a multivariate descriptor and a uni- or multivariate response block. PLS, like PCA, is a bilinear and model free method. Next to the predictive performance, the (graphical) diagnostics for model interpretation are explained. We will examine both the (essential) mathematics and the practical issues of PLS modeling, using the example to illustrate choices and results.

Author: phd@dina.kvl.dk. Updated: 09 november 2001