Comparison of Model Reliabilities from Single-Step and Bivariate Blending Methods


  • Matti Taskinen MTT Agrifood Research Finland
  • Esa Mäntysaari MTT Agrifood Research Finland
  • Martin Lidauer MTT Agrifood Research Finland
  • Timo Knürr MTT Agrifood Research Finland
  • Jukka Pösö Faba
  • Guosheng Su Aarhus University
  • Gert Pedersen Aamand NAV
  • Ismo Strandén MTT Agrifood Research Finland


genomic evaluation, reliability, single-step genomic evaluation


Model based reliabilities in genetic evaluation are compared between three methods: animal model BLUP, single-step BLUP, and bivariate blending after genomic BLUP. The original bivariate blending is revised in this work to better account animal models. The study data is extracted from the production trait evaluation of Nordic Red dairy cattle. Genotyped bulls with daughters are used as training animals, and genotyped bulls and producing cows as candidate animals. For simplicity, size of the data is chosen so that the full inverses of the mixed model equation coefficient matrices can be calculated. Model reliabilities by the single-step and the bivariate blending methods were higher than by animal model due to genomic information. Compared to the single-step method, the bivariate blending method reliability estimates were, in general, lower. Computationally bivariate blending method was, on the other hand, lighter than the single-step method.