Multivariate standardization for correcting the ionic strength variation on potentiometric sensor arrays
Abstract
Two kinds of multivariate calibration models for predicting the concentrations of K(I) and Ca(II) in natural waters from the signals of an array of potentiometric sensors were constructed. For one model, the calibration and validation samples were pre-treated by adding an ionic strength adjuster (ISA). The samples of the other model, which are called the second conditions, were not pre-treated and the model was not validated. A multivariate standardization strategy [Kennard–Stone selection and piecewise direct standardization (PDS)] was used to correct the responses of these samples. The PDS algorithm considers the arrangement of the variables, the window size and the number of the transference samples. When the conditions were optimum, there was no bias in the predictions, which were similar to those of the validated model, to which ISA had been added.