Table 6.

Prediction accuracy (correlation between phenotypes, average over 50 training-testing partitions), by trait, cluster, and model (wheat data set).

Trait	Group	BRR			BayesB
		Across groups	Interaction model	Stratified analyses	Across groups	Interaction model	Stratified analyses
E1	G1	0.453 (0.082)	0.466 (0.079)	0.455 (0.080)	0.449 (0.083)	0.467 (0.079)	0.461 (0.079)
	G2	0.573 (0.079)	0.612 (0.079)	0.613 (0.078)	0.559 (0.080)	0.604 (0.081)	0.604 (0.079)
E2	G1	0.463 (0.079)	0.463 (0.082)	0.443 (0.083)	0.453 (0.080)	0.459 (0.082)	0.440 (0.085)
	G2	0.505 (0.088)	0.488 (0.086)	0.464 (0.090)	0.501 (0.085)	0.494 (0.084)	0.462 (0.089)
E3	G1	0.410 (0.075)	0.406 (0.071)	0.396 (0.073)	0.410 (0.073)	0.408 (0.071)	0.396 (0.072)
	G2	0.362 (0.114)	0.381 (0.092)	0.373 (0.088)	0.361 (0.111)	0.376 (0.091)	0.370 (0.089)
E4	G1	0.443 (0.072)	0.460 (0.070)	0.458 (0.071)	0.442 (0.071)	0.455 (0.071)	0.454 (0.073)
	G2	0.445 (0.089)	0.487 (0.075)	0.489 (0.072)	0.442 (0.088)	0.482 (0.074)	0.481 (0.073)
Average		0.457	0.470	0.461	0.452	0.468	0.459

E1–E4 are four different mega environments.

BRR Bayesian Ridge Regression (Gaussian Prior).