SECOND-ORDER ADVANTAGE WITH DATA LOSING THE BILINEARITY IN A SINGLE SAMPLE. A NOVEL NON-BILINEAR ADAPTED PARTIAL LEAST SQUARES/RESIDUAL MODELING METHOD

SCHENONE A V; CULZONI M J; GOICOECHEA H C

Richmond

Congreso; Chemometrics in Analytical Chemistry; 2014

The most challenging data structure problem to achieve analyte quantitation from second-order data in the presence of uncalibrated components using multivariate calibration methods is the loss in bilinearity, i.e. different profiles for a component in a single sample (the shape of the profile changes with the change in the other sensor). Total synchronous fluorescence spectroscopy (TSFS) generates matrices which constitute a typical example of this kind of data. An approximation of the non-bilinear profile of the interferent can be partially achieved by modeling TSFS data with unfolded partial least squares with residual bilinearization (U-PLS/RBL). A new residual modeling for non-bilinear data is here presented, namely unfolded partial least squares with residual modeling of non bilinear data (U-PLS/RMNB). Simulated data show that the new model can conveniently handle the studied analytical problem with better performance than U-PLS/RBL. One example involving TSFS matrices illustrate the ability of the new method to handle experimental data as well: the determination of the anticancer doxorubicin in human plasma samples.