Application of robust regression in multicomponent UV spectrophotometry by direct calibration

A robust procedure, called sine function M-estimator (SFME), used in multivariate analysis, is described and utilized in multicomponent simultaneous analysis and measurement wavelength selection for a pharmaceutical system with two analytes (ephedrine hydrochloride and furaciline) by a direct calibration method. The regression results indicated that, when the noise distribution is not normal, SFME can obtain far better estimated concentration parameters than the commonly used least squares method. For the system tested, SFME yielded a four- to seven-fold improvement in prediction capability over the least squares method, together with good results in wavelength selection. The computational problems associated with SFME are discussed.