Based on the methodology for estimating breeding value for longevity of dairy cattle by random regression model, testation procedure, ZAVADILOVA, Ludmila and KASNA Eva, ISBN 978-80-7403-186-1, 2018 estimate of breeding values for Holstein cattle and comparing the results of the multi-trait procedure and estimating the breeding values for longevity based on routine practice.
To conclude, while the official method of estimating breeding values for longevity overestimates young bulls whose daughters have not yet proven their longevity, not all used linear models provide credible estimates. When comparing them, we found that the estimate of breeding values by the multi-trait model was stable. Rank correlations with routine estimates were always negative regardless of the used breeding value method. At the same time, estimation by multi-trait model converges better then random regression models and therefore the resulting breeding values are more reliable than those with random regression models that converge very poorly. These models provided rank correlations with routine estimates dependent on breeding value selection. Their results have been shown to be highly dependent on the random regression function used, so we recommend the multi-trait model for further use in the official estimate, replacing the current nonlinear regression model after implementation time.