OPTIMIZATION OF PARTIAL-LEAST-SQUARES CALIBRATION MODELS BY SIMULATION OF INSTRUMENTAL PERTURBATIONS

A critical step in partial-leasts-squares (PLS) modeling is the model optimization, Cross-validation is often applied, but in spite of its s tatistical properties, it suffers some severe shortcomings, In particu lar, cross-validation has a tendency to give overfitted models, wherea s parsimonious models should be preferred, We propose an alternative f orm of internal validation, based on the simulation of instrumental pe rturbations on a subset of calibration samples, A simple criterion is proposed for the adjustment of perturbations, The method is applied fo r the validation of nine PLS1 calibration models on industrial data se ts and compared with cross-validation and cross-validation combined wi th a randomization test. It is shown that parsimonious models can be o btained, with a good predictive power when they are applied to externa l test data.