Unfolded partial least squares/residual bilinearization combined with the Successive Projections Algorithm for interval selection: enhanced excitation-emission fluorescence data modeling in the presence of the inner filter effect.

The use of the successive projections algorithm (SPA) for elimination of uninformative variables in interval selection, and unfold partial least squares regression (U-PLS) modeling of excitation-emission matrices (EEM), when under the inner filter effect (IFE) is reported for first time. Post-calibration residual bilinearization (RBL) was employed against events of unknown components in the test samples. The inner filter effect can originate changes in both the shape and intensity of analyte spectra, leading to trilinearity losses in both modes, and thus invalidating most multiway calibration methods. The algorithm presented in this paper was named iSPA-U-PLS/RBL. Both simulated and experimental data sets were used to compare the prediction capability during: (1) simulated EEM; and (2) quantitation of phenylephrine (PHE) in the presence of paracetamol (PAR) (or acetaminophen) in water samples. Test sets containing unexpected components were built in both systems [a single interference was taken into account in the simulated data set, while water samples were added with varying amounts of ibuprofen (IBU), and acetyl salicylic acid (ASA)]. The prediction results and figures of merit obtained with the new algorithm were compared with those obtained with U-PLS/RBL (without intervals selection), and with the well-known parallel factors analysis (PARAFAC). In all cases, U-PLS/RBL displayed better EEM handling capability in the presence of the inner filter effect compared with PARAFAC. In addition, iSPA-U-PLS/RBL improved the results obtained with the full U-PLS/RBL model, in this case demonstrating the potential of variable selection.

A linear semi-infinite programming strategy for constructing optimal wavelet transforms in multivariate calibration problems.

A novel strategy for the optimization of wavelet transforms with respect to the statistics of the data set in multivariate calibration problems is proposed. The optimization follows a linear semi-infinite programming formulation, which does not display local maxima problems and can be reproducibly solved with modest computational effort. After the optimization, a variable selection algorithm is employed to choose a subset of wavelet coefficients with minimal collinearity. The selection allows the building of a calibration model by direct multiple linear regression on the wavelet coefficients. In an illustrative application involving the simultaneous determination of Mn, Mo, Cr, Ni, and Fe in steel samples by ICP-AES, the proposed strategy yielded more accurate predictions than PCR, PLS, and nonoptimized wavelet regression.