A transparent tool for seemingly difficult calibrations: the parallel calibration method.

ABSTRACT

A new easy-to-understand calibration method for the analysis of spectral data is developed. The "parallel calibration" method is logically simple and intuitive yet often provides an improvement over more complex standard calibration methods. A description of the algorithm with a technical justification for the parallel algorithm is presented, underscoring the simplicity of the approach. In addition, performance as compared to that of the standard methods of classical least-squares (CLS) and partial least-squares (PLS) regression is studied. Calibrations are carried out on a computer-generated simulation data set as well as two scientific data sets. The results show that the parallel method gives results comparable to or better than those of CLS and PLS methods in terms of mean squared error.