This zigzag pattern is counter-cyclical to that observed in the realized accuracies in the sense that when realized accuracy is up, ΔFP is down, and vice-versa.
For highly selected traits, realized accuracy of GEBV is smaller because it accounts for selection.
In Panel A, realized accuracies were normalized using a constant trait heritability that was estimated from the best fit BayesC0 analysis, and variation in realized accuracy reflects variation in the correlation between phenotype and DGV.
The largest difference between the two accuracy estimates was observed for gBLUP, where r (PEV ) under-estimated the realized accuracy when half-sib family information was used.
Finally, use of uninformed starting values for variance components led to the over-estimation of heritability, and systematic underestimation of realized accuracy.
A zigzag pattern was also present: the Bulmer effect was stronger in the generation after model updating, that is, after realized accuracy was the strongest.
The very low accuracy for females for T1, especially with BLUP, was due to strong phenotypic preselection of females based on T1; in case of extreme selection, the realized accuracy tends towards zero.
When the optimal number of PC was determined based on realized accuracy in the testing population, PCR showed a higher potential in terms of achievable accuracy that was not capitalized when PC selection was based on cross-validation.
There are large discrepancies between theoretical accuracy (e.g., by inversion of the coefficient matrix of the mixed model equations) and realized accuracy of EBV in populations under selection, where the latter is noticeably smaller [ 11].
For each trait, the realized accuracy of DGV was calculated as the pooled correlations between DGV and phenotypes in validation groups divided by the square root of trait heritability.
