Fig. 2

Heatmap of OTS (optimized training sets) graph for the USP dataset. a OTS 1 for kernel GET. b OTS 2 for kernel GET. c OTS 3 for kernel GET. d OTS 1 for kernel GWT. e OTS 2 for kernel GWT. f OTS 3 for kernel GWT. In green (green = 1 or presence of this information in the training set) we see the distribution of the information selected to form the training population, for each trait x environment and replication inside kernels, while in blue (blue = 0 or absence of the information in the TRS), we see what formed our testing set, or what remained to be predicted. The solid line that crosses all the graphs represents the genotype used as a check. The environments on the x-axis: AN.16 (Anhembi 2016), PI.16 (Piracicaba 2016), AN.17 (Anhembi 2017), and PI.17 (Piracicaba, 2017); the traits under study: EH (ear height), GY (grain yield) and PH (plant height). The kernels: GET (genotype × environment × trait) and GWT (genotype × environmental covariates × trait) are used as the base to select the information