LIMITATIONS OF EVOLVING PRINCIPAL COMPONENT INNOVATION ANALYSIS FOR PEAK PURITY DETECTION IN CHROMATOGRAPHY

Evolving principal components innovation analysis (EPCIA) is a factor analysis based method for rank estimation in ordered data sets. In thi s paper, the limitations of EPCIA when applied to the problem of unres olved two-component mixtures in liquid chromatography with multiwavele ngth UV-visible absorbance detection are examined. It is shown that th e sequence of root-mean-square prediction errors from the EPCIA models approximates the elution profile of the minor component, and that the maximum of this sequence determines the detectability of the minor co mponent. The effects of spectral correlation, chromatographic resoluti on, concentration ratio, and other factors on the performance of the a lgorithm are examined. Guidelines for predicting the utility of the al gorithm in specific cases are provided and demonstrated with experimen tal data.