this was evenmorepronounced thaninthe ‘ChokAnan’ experiment(Kienzle et al., 2011). Prediction accuracy was rough [RPDy: 2.0–2.5 (Nicolaï et al., 2007)] for b∗m (Table 7), with an error (RMSECVb* = 5.00)
representing ∼13% of the total range measured. For TA, the PLS model only allowed discrimination of low levels from maximum ones (RPDy: 1.5–2.0).