Table 5. Average percent change in prediction accuracy of 10% and 20% diversity and prediction cores vs. prediction accuracy of random cross-validation TRN20-TST80 (first four columns) and percent change in prediction accuracy between 10% diversity core vs. 10% prediction core, and between 20% diversity core vs. 20% prediction core for traits measured in one or in two environments for Mexican and Iranian collections.
| Trait-Collectiona | 10% Diversity | 10% Prediction | 20% Diversity | 20% Prediction | 10% Diversity vs. 10% Prediction | 20% Diversity vs. 20% Prediction |
|---|---|---|---|---|---|---|
| Accounting for population structure | ||||||
| One environment -Mexican collection | 8.0 | 10.3 | 2.7 | 1.2 | 2.4 | −1.6 |
| One environment -Iranian collection | 13.1 | 14.9 | 2.8 | 8.7 | 1.2 | 5.9 |
| Two environments -Mexican collection | 7.5 | 6.5 | 3.0 | 1.2 | −1.3 | −1.9 |
| Two environments -Iranian collection | 12.1 | 7.6 | 0.9 | 3.3 | −5.3 | 2.4 |
| Not accounting for population structure | ||||||
| One environment -Mexican collection | 4.1 | 4.7 | 0.9 | 0.9 | 0.6 | −0.1 |
| One environment -Iranian collection | 7.5 | 25.5 | 1.94 | 9.5 | 1.9 | 19.5 |
| Two environments -Mexican collection | 2.51 | 2.72 | 0.59 | 0.63 | 0.13 | 0.03 |
| Two environments -Iranian collection | 8.30 | 5.11 | 0.81 | 1.46 | −3.48 | 0.63 |
a
Traits evaluated in one environment when accounting for population structure (from Table 2), traits evaluated in two environments when not accounting for population structure (from Table 4), traits evaluated in one environment when not accounting for population structure (from Table A1), traits evaluated in two environments when not accounting for population structure (from Table A3).