Abstract
When high-performance liquid chromatography（HPLC）is applied to multicomponent medicines, quality is quantified by the use of waveform patterns and the Mahalanobis-Taguchi system offers an effective means of making pass/fail decisions. Incorrect decisions have posed a problem, however, because of the wide tolerance ranges of multicomponent medicines, and because of external factors that cause unavoidable scatter in the results of HPLC analysis. We have therefore studied methods of creating a more robust unit space by building scatter into the unit space so that the pass/fail decisions are more immune to scatter and tolerance effects. Two such methods were studied: in one, the unit space was created from data that included analytic scatter generated by random numbers; in the other, principal component analysis and Cholesky decomposition were applied to data having the same factor-to-factor correlation coefficients and corresponding to the passing tolerance range. These methods were shown to be capable of creating unit spaces with improved discriminatory power that were robust with respect to the above unavoidable sources of error.
https://ror.org/05jk51a88 
