Single-trait and multi-trait prediction of breeding values for show-jumping performance of horses in the Czech Republic

Genetic parameters for show-jumping performance of horses in the Czech Republic were estimated from 483,303 observations of 17,542 horses recorded between 1991 and 2010. The results from events did not have normal distributions. Data were analysed with a least-squares method (GLM/SAS), and genetic parameters were estimated through a Gibbs sampling method. The statistical model included fixed effects for sex, year of the event, level of difficulty of the event and random effects for rider, permanent environment and an additive genetic effect. Six transformations of the data were tested, and the most suitable evaluation was chosen on the basis of lowest residual variance, highest heritability and closest approximation to normal distributions of residuals and breeding values. By these criteria, the best evaluation was accomplished with the shifted Blom-normalised rank for penalty points. For comparison, breeding values were predicted with a single-trait and multi-trait animal model. In the multi-trait model, each record was assigned to one of three traits on the basis of the difficulty of the performance event (i.e., fence height 90-110 cm defined the first trait, 120-135 cm the second trait and 135-150 cm the third trait). The heritability estimates of show-jumping performance were 0.07 for the single-trait model and 0.07, 0.10 and 0.16 for the multi-trait models. Relative breeding values and relative commercial values of the horses were calculated. Both had a normal distribution, and positive genetic trends were estimated for the relative breeding values.