Although chemometrics has become widely used recently for analyzing experimental chemical data, there exist only a few instructions for the proper usage of chemometrics other than those in some introductory books. As the sixth step of chemometrics calculations with Microsoft Excel (Excel), the partial least-squares (PLS) regression (PLSR) is performed on worksheets. The worksheets were prepared for generating the spectra of model calibration samples and unknown samples, obtaining latent variables by PLS1 (single objective variable model) and by PLS2 (multi objective variable model), and constructing quantitative model by PLSR. The quantitative performances of PLSR and PCR were compared by using the generated unknown spectra. PLSR model with PLS1 has good performance with a small number of factors for the chemical species using objective variable. PLSR modeling with PLS2 computes stable results for each chemical species. These results indicate that PLSRs are superior to PCR, when a small number of factors are used. However, it is not obvious which method produces the best quantitative model. For getting the best model, it is necessary to compare the methods using various factors.
2014 Society of Computer Chemistry,