LAUSR

Abstract Multivariate curve resolution (MCR) methods as MCR-ALS, ReactLab, the peak group analysis and SVD-based hard-modeling methods differ in their algorithms and the underlying optimization procedures. These differences include variants in the implementation of the algorithms and a differing weighting of the constraints. Depending on the MCR method different computational results can be obtained for the same data set. The area of feasible solutions (AFS) comprises all possible outcomes of MCR methods. It represents all nonnegative factors of a given spectral data set. It therefore offers an unbiased view of the problem. In a comparative study we present within the AFS the various MCR results for a model data set and for experimental FTIR data. For the model data we observe that the spread of the MCR results correlates with the so-called purity of the spectral data.