PLS-DA and ECVA use all spectral regions for analysis however
some parts are redundant and should be discarded to gain more
predictive models. The interval PLS-DA (iPLS-DA) algorithm selects
the best single interval of the spectrum, based the performance of
PLS-DA models built on a series of user-selected intervals. This is
an attractive and simple approach to wavelength selection, but improper
selection of the size of the interval used in iPLS-DA can corrupt
the predictive performance from iPLS-DA regression model.
The same is applicable for ECVA and the respective model is called
iECVA.