Fig. 4

Relationship between training set size and predictive ability of models for grain yield (YLD), grain moisture (GM), and percentage of oil (OIL), calibrated with TRS obtained by various optimization methods. The plot shows the average predictive ability across iterations of the training set optimization and repetitions of the gradient boosting machine model for two different combinations of candidate and test set years. The x-axis represents the size of the training set as a percentage of the candidate set. Error bars indicate the standard error of the mean. The gray horizontal line represents the average predictive ability achieved when using the entire candidate set to calibrate the prediction models and the gray area around it shows the standard error of the mean