TABLE 8.

G2F_2016 data. Count of the number of times a model was better than another in terms of Normalized Root Mean Squared Error (NRMSE), both by environments and by traits. Prediction accuracy in terms of NRMSE. b0 denotes the absolute values of the intercept and b denotes the absolute value of 1-slope. Under an ideal model both b0 and b should be equal to zero. Relative efficiency (RE) or each model in percentage was computed regarding model M0 without environmental covariates. When the percentage is positive there is a gain in prediction accuracy regarding M0, while when the percentage is negative there is a loss in terms of prediction accuracy of any model regarding M0.

Model	Env		Trait		NRMSE	NRMSE	Cor	Cor	b0	b0	b	b
	Won	%	Won	%		RE (%)		RE (%)		RE (%)		RE (%)
	Models		Models
M0	447	31.04	23	28.75	4.312	0.00	0.493	0.00	0.479	0.00	0.089	0.00
M1	502	34.86	30	37.50	6.009	−28.24	0.491	−0.34	0.088	444.83	0.098	−9.21
M2	440	30.56	17	21.25	3.522	22.44	0.471	−4.45	1.278	−62.48	0.011	725.99
M3	558	38.75	32	40.00	3.399	26.85	0.492	−0.22	0.157	205.09	0.101	−11.68
M4	470	32.64	17	21.25	3.799	13.51	0.400	−18.78	0.880	−45.50	0.119	−25.45
M5	560	38.89	35	43.75	3.573	20.70	0.492	−0.26	0.381	25.98	0.093	−4.61
M6	593	41.18	38	47.50	4.154	3.81	0.492	−0.13	0.084	467.90	0.104	−14.20
M7	477	33.13	27	33.75	4.211	2.39	0.401	−18.73	0.889	−46.07	0.118	−24.74
M8	472	32.78	28	35.00	3.552	21.38	0.401	−18.69	0.796	−39.78	0.113	−21.29
M9	297	20.63	3	3.75	5.933	−27.32	0.430	−12.69	0.667	−28.09	0.003	2,916.95
M10	492	34.17	22	27.50	4.925	−12.45	0.406	−17.54	0.624	−23.17	0.151	−40.99
M11	479	33.26	30	37.50	6.058	−28.82	0.444	−9.87	1.014	−52.72	0.041	116.41
M12	542	37.64	33	41.25	4.589	−6.03	0.457	−7.39	1.381	−65.29	0.005	1,645.10
M13	618	42.92	43	53.75	3.474	24.13	0.492	−0.18	0.604	−20.59	0.093	−4.79
M14	604	41.94	42	52.50	3.436	25.50	0.492	−0.20	0.565	−15.12	0.092	−3.58