Validation of the evaluation of longevity by weighted analysis

We applied the Interbull validation methodology, a widely accepted method in animal breeding, to assess novel weighted and nonweighted repeatability models for the prediction of breeding values for longevity in Czech Holstein cattle. The population included in the estimation also contained 58 704 animals with genotypes. Performance records from 1 055 814 cows in the full dataset and from 831 995 cows in the trimmed dataset were used for evaluation. Both linear models included effects of herd-year-period, animal, permanent environment, and correction for milk production of individual cows in relation to herd average lactation curve, differing only in the use of weighs. The average reliability of the validation bulls increased in the full dataset from 0.85 without weight to 0.91 in the model with weight. This increase was also apparent in the trimmed dataset (from 0.42 to 0.50). Both models showed considerable inflation of genomic breeding values (GEBVs) by Interbull validation and did not manifest distinct benefits supporting their use in the routine evaluation of Czech Holstein cattle in the Czech Republic.