Multivariate Statistical Analysis of Spectroscopic Data

This paper focuses on the application and comparison of Principal Component Analysis (PCA) and Independent Component Analysis (ICA) using two generic artificially created datasets. PCA and ICA are assessed in terms of their abilities to infer reference spectra and to estimate relative concentrations of the constituent compounds present in the analysed samples. The results show that ICA outperforms PCA and is able to identify the reference spectra of all the constituent compounds. On the other hand, PCA fails to identify one of the constituent compounds for the first dataset. Also, ICA estimates relative concentrations of all the constituent compounds present in both datasets much more accurately than PCA.