The application of the method for the simultaneous spectrophotometric determination of the divalent ions of copper, manganese and zinc for analysis of a pharmaceutical formulation of polivitaminic/polimineral is reported. This method uses 4-(2-pyridylazo) resorcinol (PAR) and multivariate calibration and was optimized using the successive projections algorithm (SPA) and a genetic algorithm (GA) for choosing the best series of wavelengths for analysis. Thus the construction of calibration models based on multiple linear regression (MLR-SPA and MLR-GA respectively) was made possible. The results obtained were compared with models based on latent variables, principal component regression (PCR) and partial least square regression (PLS) through the criterion of the root mean square error of prediction (RMSEP). All the methods presented even performance but the RLM methods are simpler since they require a smaller number of wavelengths and are easier to interpret than those based on latent variables.
variable selection; multivariate calibration; metals; drugs analysis; spectrophotometry
