Multivariate chemometric techniques were used t o classify alcoholic distillates and t o develop a typification model for Galician liquors, o n the basis of percentage data obtained from nine chromatographic peaks. By using the Bayesian model, the probability of a genuine Galician liquor being rejected is 0.1 1 and that of a false one being accepted is practically nil. Partial least squares was used as a modelling method, taking the liquor category as response variable. This method enables a confidence interval (95%) t o be constructed that does not include any of the other distillates.

No abstract

No abstract

No abstract

The ability of the least median squares (LMS) method in detection of several straight segments in experimental continuous curves is tested using chronopotentiometric data. As a common experimental system, the chronopotentiometric reduction of Fe(CN)a-on a Pt electrode is used for this purpose. LMS allows for the correct estimation of transition times and then, for the calibration of the system. LS calibration without outliers detected by LMS assures a better prediction and therefore better calibration quality. beCduSe the standard error of estimation is 32.43% less than one obtained with the whole data.