Data analysis by iPLS-DA was performed on both original data
and those pre-processed by EMSC method. Since better results
were obtained by the EMSC pre-processed data, the results obtained
by this data will be discussed here. An important issue in
iPLS is selecting the number of intervals, which was found here
by trial and error. Among the different number of segments examined,
an iPLS-DA model of 20 equidistant sub-intervals (resulted in
lower prediction error) was selected. For each sub-interval, the