A general view of multi-way data processing and its advantages

OLIVIERI, A. C.

Saratov

Simposio; XII Winter Symposium on Chemometrics; 2020

Universidad de Saratov

Multi-way data processing and calibration display outstanding properties from the analytical chemistry perspective. For example, by measuring chromatographic-spectral data matrices, it is possible to quantitate analytes in complex samples, without the need of a large training set. Only a few pure analyte standards are needed; after measuring their matrix data and joining them with those for the unknown sample, a multiway calibration model mathematically separates the analyte contribution from those of the interferences. This makes it unnecessary to apply sample pre-treatment or clean-up procedures, to achieve baseline resolution of every analyte peak or to correct for background signals. Chromatographic protocols become simpler, isocratic, cheaper, faster and, perhaps more importantly, greener. All these benefits stem from the second-order advantage, which is inherent to data second-order and beyond, i.e., matrices, three-dimensional arrays, etc.In this lecture, the following topics will be covered: (1) data structures, (2) multi-way models and (3) experimental applications. Instrumental data will be classified according to the so-called ?order? (number of modes for a single sample) and ?ways? (number of modes for a sample set). More importantly, multi-way data need to be classified as multi-linear or non-multilinear, and in this case according to the cause for the multi-linearity loss. It will be shown that this latter classification scheme is highly helpful for selecting the correct data processing model. A number of experimental reports will serve to illustrate the various data properties and the corresponding multi-way models employed for calibration. The examples include the use of matrix excitation-emission fluorescence spectroscopy and of liquid chromatography with multivariate spectral detection. Some cases involving four- and five-way calibrations will be shown, where a kinetic evolution or a pH gradient are implemented as additional instrumental modes.