Validation of alternative models in genetic evaluation of racing performance in North Swedish and Norwegian cold-blooded trotters.

This research evaluates different genetic models to predict racing performance in North Swedish and Norwegian cold-blooded trotters. The study compared the current bivariate model focusing on racing status and earnings with a univariate model and a threshold-linear animal model.

Methods and Approach

• The authors approached the validation of different ways of estimating the breeding values of performance in trotters, specifically North Swedish and Norwegian cold-blooded trotters.
• The current genetic evaluation method, which considers both the racing status (RACE) and earnings (EARN) of the horses, was compared with two alternatives: a threshold-linear animal model and a univariate model that only focuses on the performance trait.

Comparison Techniques

• The distinct methods were compared using a cross-validation process of standardized earnings. Mean Squared Errors of Prediction (MSEP) and the correlation between the phenotype (the observed characteristics) and the estimated breeding value (EBV) were utilized as the criteria for comparison.

Results and Findings

• Despite potential influences of selection, the study found a relatively high heritability estimate for EARN in the univariate analysis, meaning earnings were a strong predictor of genetic performance.
• When comparing the two models, the genetic trend estimate for earnings was higher in the bivariate model than in the univariate model. This was attributed to the considerable size of estimated heritability of RACE and its high correlation with EARN (approximately 0.8).
• The study pointed out that RACE tends to reflect ancestry rather than the horseracing performance of the horse itself. Because of this, using RACE in the genetic evaluation may overstate the genetic trend of EARN due to duplicated pedigree information.

Recommendations

• Despite the issue with potential inflation of genetic trends of EARN, the authors recommend sticking with the current bivariate model. The authors reason that the bivariate model has a higher predictive ability and improves the ranking of animals within a year-class. It also addresses the inability to differentiate between models for genetic trend.