Random augmented inverse least squares

Abstract

Calibration is an integral part of quantitative chemical analysis which can be followed in either a classical or inverse style. While calibration methods have been investigated since the early days of analytical chemistry, they have been periodically refreshed and modified over the years. As a part of this ongoing progress, a novel generic calibration scenario has been implemented to treat partial knowledge of the concentration profiles, offering an alternative approach to performing ILS through a regularization process. The primary feature of the proposal is to design a calibration set supplementing the analyte’s concentration with artificial vectors of random numbers to compensate for any potential interferent contributions. In this way, the new inverse calibration, titled Random Augmented Inverse Least Squares (RAILS) uses only the concentration of the analyte and the number of n-1 pseudo-components with random concentration profiles must be optimized. Monte Carlo Cross-Validation (MCCV) was employed to optimize the urgent number of dummy concentration vectors required during RAILS and it was tested across several simulated and real experiments from different spectroscopic schemes.