LAUSR

This paper presents a simple and sensitive method for the simultaneous determination of methyl paraben (MP) and phenol (PO) based on the application of successive projections algorithm (SPA) to the first derivative spectra (200–350 nm). SPA is used for variables selection in order to obtain multiple linear regression (MLR) models using a small subset of wavelengths. The starting vector and the number of variables are optimized and the best variables are selected according to the sequence of projection operations on the spectral data matrix of the calibration set. Principal component regression and partial least squares models are also developed for comparison. The best models are found to be SPA-MLR using seven wavelengths from the first-derivative spectra with a root-mean-square error of prediction (RMSEP) of 0.08 for MP and eight wavelengths with RMSEP of 0.31 for the determination of PO. The accuracy of the proposed method is confirmed by spiked recovery test on cosmetic samples with satisfactory results (86–110%). Analysis results of the cosmetic samples are also statistically compared with those obtained from the HPLC method, showing no significant difference regarding accuracy and precision. The results indicate the potential of SPA-MLR and derivative spectrophotometry for rapid and sensitive analysis of cosmetic samples.