The effect of milk sampling procedures on the robustness of genetic evaluation in dairy cattle

The study was divided into three separate sections. Firstly the influence of selection is investigated by means of simulating a population including selection in the model and then simulating data with out selection. In the second section the simulation process was validated by determining how well the breeding values and the resulting variance components were predicted under DFREML procedures. This was done by the calculation of the fixed effect, variance components and product moment correlation between the estimated breeding values and the simulated breeding values. In the third section a multiple trait analysis was done in order to determine the relationship between the single and composite sample. The effect of selection showed a substantial overestimation of the fixed effect and the resulting variance component estimates. Various authors (Hagger, 1991; Tuchsceherer & Herrendërfer, 1998, van der Werf, 1990) supported these results and this effect can be attributed to fact that all information on all animals in the analysis should be available to compensate for selection. The product moment correlation between the true genetic component and the breeding values determined by DFREML as well as the estimated variance components confirmed that the simulation process and the model used to simulate data can be applied to simulate data. These results were used in a multiple trait analysis to evaluate the difference between a single and composite sample analysis. Results from multiple trait analysis showed a high genetic correlation between the breeding values of the composite and single sample. This indicates that the same genes are responsible for both traits and that it is indeed possible to make use of a single sample instead of a composite sample.