Table 2.

Dataset 1 EYT 2013–2014

	Models and methods					Models and methods
	BRR	GBLUP	GK	PK	SK	BRR	GBLUP	GK	PK	SK
Scenario	Without G × E (WI)					With G × E (I)
DTHD
MT	15.24	15.20	14.24	14.37	21.19	12.77	12.58	11.60	12.43	20.67
MT_P	14.18	14.18	13.36	13.53	19.32	12.18	11.97	11.06	11.73	18.96
DTMT
MT	13.64	13.61	12.96	13.03	18.44	11.79	11.63	10.73	11.24	18.38
MT_P	12.28	12.30	11.68	11.80	16.37	10.90	10.74	9.92	10.31	16.39

Average mean squared error (MSE) prediction across environments for five model-methods: BRR, Bayesian ridge regression; GBLUP, genomic best linear unbiased predictor; GK, Gaussian kernel; PK, polynomial kernel; SK, sigmoidal kernel without G × E (WI) and with G × E (I) for two scenarios (MT and MT_P), four environments (Bed5IR, EHT, Flat5IR, LHT), and two traits (DTHD, days to heading and DTMT, days to maturity). Boldface indicates model-method with the lowest MSE for each scenario.