EIGENSTRUCTURE TRACKING ANALYSIS FOR REVEALING NOISE PATTERN AND LOCAL RANK IN INSTRUMENTAL PROFILES - APPLICATION TO TRANSMITTANCE AND ABSORBENCY IR SPECTROSCOPY

The heteroscedasticity in mixture-designed three-component absorbance IR spectra is revealed by means of eigenstructure tracking analysis (E TA). A moving window containing neighbouring spectral wavelengths for all samples defined submatrices that were decomposed by means of singu lar-value decomposition. Starting from window size two, the ETA proced ure continues with window size three, four etc. until the window size exceeds the highest local chemical rank by one. The last evolving eige nvalue for this window provides the structure of the local noise. Due to closure, the maximum window size needed for the three-component spe ctra analysed in this work was three. Eigenstructure tracking analysis reveals that the amount of noise in absorbance IR spectra depends exp onentially on signal variance, reflecting the transformation from tran smittance to absorbance IR spectra. Transmittance IR spectroscopy is f ound to show almost homoscedastic noise behaviour.