The determination of the number of component in the PLS-DA model based on balanced error rate (BER)? what is the different between BER and predicted residual sum of squares (PRESS) in PLS?

Regards

The determination of the number of component in the PLS-DA model based on balanced error rate (BER)? what is the different between BER and predicted residual sum of squares (PRESS) in PLS?

Regards

Hi @aurum,

How BER is calculated: The classification error rate for each class is calculated. These values are then averaged to get a balanced error rate which gives all classes the same weight regardless of the number of samples in them. It is between 0 and 1 and a lower value indicates a rather accurate model. Tuning for BER ensures all classes are equally accounted for in the model fit.

PRESS: The residual sum of squares for the samples **which were not used to fit the model**. It can take any non-negative value. A lower value indicates a good fit. Tuning for PRESS helps avoid over-fitting.

Hope it helps.

Al