The design of calibration in near infra‐red reflectance analysis by clustering

Abstract

Abstract A procedure for selecting samples to use in multivariate calibration, inspired by a paper by Zemroch, is proposed. The method developed is designed for multicollinear data and is based on cluster analysis. It selects among availabe spectra the samples to submit to chemical analysis and calibration. The method is tested on an example from NIR spectroscopy of biscuit dough pieces.