Selecting and visualizing the spectral variability relevant for sample classification using principal component analysis
In this work we present a simple procedure based on principal component analysis (PCA) to reconstruct a measured spectrum by selecting the portion of its total variance of interest. We also provide an approach to the understanding of the results provided by PCA, which may be useful for spectroscopists that are unfamiliar with PCA. Our proposed procedure is useful for studying the correlations between the energy channels of a given spectrum and it also leads to the construction of a new filtering method. Its potential is shown by applying it to X-ray emission and X-ray resonant Raman scattering spectra. Since the proposed procedure is independent of the spectra under study it can be a useful tool for addressing and interpreting the covariance structure of a measured spectrum for any spectroscopist.