MULTIVARIATE PREDICTION BY USING CONVENTIONAL AND DERIVATIVE PARTIAL LEAST-SQUARES STATISTICS IN A COMPLEX CHEMICAL-SYSTEM - TERNARY METALS MIXTURE RESOLUTION

The quantitative prediction abilities of Partial Least Squares methods (type 1 and 2) for analysis of conventional and derivative absorption spectra are compared. The influence of the band width and of the spectral overlapping on the capacity of prediction of PLS in both cases are described and ternary mixtures of metals are resolved. To obtain analytes with adequate spectral characteristics the resolution of the metals has been accomplished by using the reaction with a cyclic hydroxamic acid and extraction into methyl isobutyl ketone (MIBK). Significant advantages have been found by application of differentiation techniques in combination with PLS-1 method.