REALISTIC SIMULATIONS OF HIGH-PERFORMANCE LIQUID-CHROMATOGRAPHIC ULTRAVIOLET DATA FOR THE EVALUATION OF MULTIVARIATE TECHNIQUES

In this paper it will be shown that real high-performance liquid chromatographic-ultraviolet (HPLC-UV) data do not perfectly obey the exact linear model assumed in most multivariate techniques. These small model errors may have an influence on the performance of multivariate techniques, especially on the principal component analysis step in which a reduced multivariate space is defined. In order to obtain a more realistic picture of the performance of multivariate techniques in practice, simulations of peak clusters are proposed that make use of real peak systems of pure analytes. It wilt be shown that these realistically simulated data contain the same small model errors that can be observed for real HPLC-UV data.