Fig. 4.

Evaluation of 13 statistical methods for predicting three coffee traits—production of coffee beans, incidence of coffee leaf rust and yield of green beans—in two Coffea canephora populations (Intermediate and Premature) evaluated in two locations (FEM and FES). Refer to the “Genomic prediction methods” section for an overview of the methods compared. Prediction accuracy was computed as the Pearson correlation between the predicted and adjusted phenotype in test samples. Cross-validation were performed using a Replicated Training-Testing. In each replication, 80% of the data were assigned to train the model and 20% remaining to validate the model. The procedure was replicated 30 times