Usability of different genetic evaluation validation tests in a population subjected to a strong genomic selection and in testing the single-step genomic evaluations

Abstract

Interbull trend validation tests II and III and the dGEBV prediction test were applied on genetic and genomic evaluations of production traits in Nordic Red Dairy cattle evaluations. For all tests, both the animal model and the single-step GTBLUP evaluations were applied to a control model and to a faulty model where the calving age effect was removed. The trend test II was able to detect faulty models, while the test III had insufficient power because of only few bulls returning to service. Both trend validation tests seemed to work equally well for genetic and genomic evaluations. The dGEBV prediction test indicated problems in both genomic evaluation models, although results for inflation and bias were worse with the faulty model. Given the good properties of test II, an alternative approach was proposed for the calculation of daughter yield deviations. The approach is based implicitly on MME equations and requires only the estimates of breeding values and pedigree.