Another example is to penalise a poor mass accuracy, by multiplying it by a negative factor.

begin{wrapfigure}{r}{0.40textwidth}
vspace{-5pt}
    includegraphics[width=0.4textwidth]{discriminant.png}
    vspace{-20pt}
end{wrapfigure}

Such corrected factors are then summed up and gives a discriminant score. This is done for each spectrum and afterwards a histogram is made of these discriminant scores.