ELIMINATION OF UNINFORMATIVE VARIABLES FOR MULTIVARIATE CALIBRATION

A new method for the elimination of uninformative variables in multiva riate data sets is proposed. To achieve this, artificial (noise) varia bles are added and a closed form of the PLS or PCR model is obtained f or the data set containing the experimental and the artificial variabl es. The experimental variables that do not have more importance than t he artificial variables, as judged from a criterion based on the b coe fficients, are eliminated. The performance of the method is evaluated on simulated data, Practical aspects are discussed on experimentally o btained near-IR data sets, It is concluded that the elimination of uni nformative variables can improve predictive ability.