Lars Nørgaard, C. Ridder
1994
Citations
0
Influential Citations
6
Citations
Journal
Talanta
Abstract
A multivariate approach to the quantitative determination of 2-hydroxybenzaldehyde (salicylaldehyde), 3-hydroxybenzaldehyde, and 4-hydroxybenzaldehyde in their mixtures is described. The method is based on second order data generated in a flow injection analysis system with a pH gradient and photodiode-array detection. Each injection gave rise to an 89 (times) x 101 (wavelength) matrix, containing both the acidic and the basic characteristics of the sample injected. A least-squares algorithm based on Lambert-Beers law was used for the prediction of concentrations in unknown samples. No assumptions concerning the qualitative mixture composition of the hydroxybenzaldehydes were necessary to perform concentration predictions. The following four data types were used in the least-squares modelling: (1) unfolded raw data, (2) acidic spectra, (3) basic spectra, and (4) first spectral derivative of the raw data. The prediction errors obtained were comparable to literature results. A graphic method, based on the model residuals for detecting erroneous samples, was developed.