Heterogeneous Variance Adjustment in Across- Country Genetic Evaluation with Country-Specific Heritabilities

Correct ranking of animals in across-country genetic evaluation requires careful design of both, the model for breeding value estimation and the adjustment for heterogeneous variance. The Nordic across-country genetic evaluation for yield traits of Danish, Finnish and Swedish red dairy cattle is based on a multiple-trait random regression test-day model, which considers observations from different countries as different traits. This was supported by analysis of variances, which yielded for the same biological traits differences in heritabilities across countries (Lidauer et al., 2009). However, for the breeding value estimation it was decided to impose a genetic correlation of unity and to have the same genetic variances across countries. Both conventions may depart somewhat from the true genetic nature of the breeding population but enhance genetic comparison of animals across the three countries. The genetic correlation of unity is modelled in the breeding value estimation model, whereas the same genetic variances can be secured through adjustment for heterogeneous variance. Heterogeneous variance is accounted by a multiplicative mixed model approach (Meuwissen et al., 1996). The method solves the model for breeding values and the model for estimating the heterogeneity of variance simultaneously with the aim to scale phenotypic observations until re-estimated residual variances are the same for all the strata. Because variance ratios remain unaffected by the method, it is possible to calibrate the heterogeneous variance adjustment method to yield the same genetic variances across countries, given the genetic variances can be estimated from the estimated breeding values (EBV). Currently, genetic variances are standardized across the countries based on standard deviations (SD) of the bull breeding values. In this study we explored the usefulness of estimating the genetic variances from cows’ EBVs applying a full model sampling method and using these estimates for calibrating the heterogeneous variance adjustment method to obtain homogeneous genetic variances across the countries.