Evaluation of empirical chemometric models for calibration and classification

A fundamental task in science and technology is searching for relationships between data sets, e. g., modeling a PROPERTY y by one or several VARIABLES x. Often a desired property cannot be determined directly or only with high cost. In contrary some x-data may be easily available. Depending on how well known and how strictly defined the relationship between x and y is, we can distinguish different levels of creating and applying models that predict y from x [1].