Table 8.

The maximum and minimum difference of the predictive ability between ELPGV and other methods in different sample size of simulation

Method	100	200	300	400	500	599
	Predictive ability	Predictive ability	Predictive ability	Predictive ability	Predictive ability	Predictive ability
ELPGV	0.6055 ± 0.0220	0.7175 ± 0.0159	0.7559 ± 0.0085	0.7189 ± 0.0070	0.7378 ± 0.0061	0.7034 ± 0.0064
BayesA	0.5486 ± 0.0256	0.6827 ± 0.0170	0.7200 ± 0.0085	0.7034 ± 0.0073	0.7231 ± 0.0069	0.6865 ± 0.0068
BayesB	0.6007 ± 0.0237	0.7075 ± 0.0170	0.7502 ± 0.0084	0.7158 ± 0.0069	0.7344 ± 0.0060	0.6985 ± 0.0065
BayesCπ	0.5334 ± 0.0215	0.6215 ± 0.0266	0.7271 ± 0.0100	0.7069 ± 0.0076	0.7354 ± 0.0060	0.6955 ± 0.0064
GBLUP	0.5266 ± 0.0216	0.4785 ± 0.0243	0.5655 ± 0.0151	0.5511 ± 0.0143	0.5934 ± 0.0095	0.5505 ± 0.0096
Maximum difference	0.0789 (15.0%)	0.2390 (49.9%)	0.1904 (33.7%)	0.1678 (30.4%)	0.1444 (24.3%)	0.1529 (27.7%)
Minimum difference	0.0048 (0.8%)	0.0100 (1.4%)	0.0057 (0.8%)	0.0031 (0.4%)	0.0024 (0.3%)	0.0049 (0.7%)

ELPGV is the ensemble learning based on BayesA, BayesB, BayesCπ and GBLUP