Impact of Using Reduced Rank Random Regression Test-Day Model on Genetic Evaluation

The development of genetic evaluations on dairy traits based on individual test-day records represents a major computing challenge due to the number of parameters in the model and the number of records to analyse. To reduce computer requirements, we proposed to use reduced rank test-day models where the smallest eigenvalues of the covariance matrix of random effects (genetic, permanent environment and herd-year) are set to zero. Different levels of reduction were tested. The model with 4 genetic, 4 permanent environment and 2 herd-year effects including heterogeneous herd-year residual variance led to only minor changes in estimated breeding values compared with the full rank model.