Comparative Study of the PLSR and PCR Methods in Laser-Induced Breakdown Spectroscopic Analysis

Abstract

Principal components regression (PCR) and partial least squares regression (PLSR) are the two most powerful and widely used chemometric regression methods applied in laser-induced breakdown spectroscopy (LIBS) studies. In this work, we compared the analytical merits (accuracy and precision) of applying the PCR and PLRS algorithms to identical LIBS data sets. A set of stainless steel samples with multielemental concentration data was studied for this work. A detailed study of the PCR and PLSR coefficients and the corresponding spectra used to generate them was carried out to understand the role of the two algorithms in the evolution of the regression coefficient signal data. Based on this understanding, a few guidelines were proposed in this work for selecting PCR or PLSR depending on the analytical situation. This work identifies the situations wherein the PCR and PLSR application results are analytically equivalent and where one approach is superior over the other for LIBS data analysis.