For a PCA or PLS model, it is necessary to determine the model complexity or the number of components. The use of a correct number of components is analogous to the model being fitted to the data in the best way possible. Too few components will under-fit the model, which means that important variation still remains in the data. Too many components will instead over-fit the model, meaning that variation accounting for noise will be brought into the model.

For PCA and PLS models, Evince automatically detects the number of components by default. This detection is based on cross-validation, where the Q2X or the Q2Y values should be above zero for the component to be significant. It is also possible to give a user-defined number of components to be calculated.