Table 7. Wheat data.

Method	Environment–trait	Correlation			MSEP
		Mean	SE	Ranka	Mean	SE	Rank
	Bed2I-DTHD	0.93	0.03	2	4.82	1.60	2
	Bed2I-GNVI	0.79	0.08	1.5	6.8E-05	0.00	2
	Bed2I-GRYLD	0.64	0.12	1	0.05	0.01	1
	Bed2I-PTHT	0.60	0.18	1	26.55	11.90	2
	Bed5I-DTHD	0.76	0.13	2	17.61	7.63	1
U	Bed5I-GNVI	0.60	0.20	1	9.7E-05	0.00	2
	Bed5I-GRYLD	0.35	0.32	2	0.23	0.09	2
	Bed5I-PTHT	0.46	0.16	3	11.90	3.40	3
	Drip-DTHD	0.95	0.02	1	2.66	0.81	1
	Drip-GNVI	0.68	0.21	1.5	0.00	0.00	2
	Drip-GRYLD	0.67	0.17	1	0.13	0.05	1
	Drip-PTHT	0.69	0.08	1	22.68	10.21	1
	Ave	0.68	0.14	1.50	7.22	2.98	1.67
	Bed2I-DTHD	0.95	0.01	1	4.44	0.57	1
	Bed2I-GNVI	0.79	0.01	1.5	6.3E-05	0.00	2
	Bed2I-GRYLD	0.60	0.04	2	0.06	0.00	2
	Bed2I-PTHT	0.56	0.06	2	28.24	3.70	2
	Bed5I-DTHD	0.79	0.04	1	16.51	2.37	1
D	Bed5I-GNVI	0.66	0.06	2	8.4E-05	0.00	2
	Bed5I-GRYLD	0.38	0.09	1	0.22	0.02	1
	Bed5I-PTHT	0.47	0.06	2	11.86	1.10	2
	Drip-DTHD	0.94	0.01	2	4.34	0.53	2
	Drip-GNVI	0.68	0.06	1.5	0.00	0.00	2
	Drip-GRYLD	0.59	0.06	3	0.14	0.02	2
	Drip-PTHT	0.62	0.03	2	23.83	3.14	2
	Ave	0.67	0.04	1.75	7.47	0.95	1.75
	Bed2I-DTHD	0.94	0.05	3	17.37	6.94	3
	Bed2I-GNVI	0.33	0.21	3	0.00	0.00	2
	Bed2I-GRYLD	0.58	0.18	3	0.07	0.01	3
	Bed2I-PTHT	0.56	0.23	3	32.70	14.42	3
	Bed5I-DTHD	0.78	0.14	3	30.94	8.83	3
S	Bed5I-GNVI	0.46	0.25	3	0.00	0.00	2
	Bed5I-GRYLD	0.38	0.33	3	0.24	0.09	3
	Bed5I-PTHT	0.41	0.18	1	9.88	3.18	1
	Drip-DTHD	0.93	0.05	3	7.27	2.56	3
	Drip-GNVI	0.43	0.14	3	0.00	0.00	2
	Drip-GRYLD	0.55	0.18	2	0.17	0.08	3
	Drip-PTHT	0.61	0.16	3	28.84	12.68	3
	Ave	0.58	0.17	2.75	10.62	4.07	2.58

Mean and SE of the estimated correlations and MSPE from the 10-fold cross-validation CV1. The BMTME model was fitted using unstructured (U), diagonal (D), and standard (S) variance–covariance matrices. Environment (Bed2I, Bed5I, Drip)–trait [days to heading (DTHD), GNDVI, grain yield (GRYLD), and plant height (PTHT)] combination. Method stands for the three variance–covariance matrices used with the BMTME.

^aSince three BMTME models are fitted (unstructured, diagonal, and standard), the values of the ranks ranged from 1 to 3, and the lower the values, the better the prediction accuracy. For ties, we assigned the average of the ranks that would have been assigned had there been no ties.