Combining Threshold, Thurstonian and Classical Linear Models in Horse Genetic Evaluations for Endurance Competitions

Abstract

The racing time and rank at finish traits are commonly used for endurance horse breeding programs as a measure of their performance. Even so, given the nature of endurance competitions, many horses do not finish the race. However, the exclusion of non placed horses from the dataset could have an influence on the prediction of individual breeding values. The objective of the present paper was to develop a multitrait model including race time (T), rank (R) and placing (P), with different methodologies, to improve the genetic evaluation in endurance competitions in Spain. The database contained 6135 records from 1419 horses, with 35% of the records not placed. Horse pedigree included 10868 animals, with 52% Arab Horses. All models included gender, age and race effect as systematic effects and combined different random effects beside the animal and residual effects: rider, permanent environmental effect, and interaction horse-rider. The kilometers per race was included as a covariate for T. Heritabilities were estimated as moderately low, ranging from 0.06 to 0.14 for T, 0.09 to 0.15 for P, and 0.07 to 0.17 for R, depending on the model. T and R appeared mostly as inverse measures of the same trait due to their high genetic correlation, suggesting that T can be ignored in future genetic evaluations. P was the most independent trait from the genetic correlations. The possibility of simultaneously processing the threshold, Thurstonian and continuous traits has opened new opportunities for genetic evaluation in horse populations, and much more practical genetic evaluations can be done to help a proper genetic selection.

Impact of the event effect in genetic evaluation for ranking traits in horses

“…The simulation was designed in the usual heritability ranges in ranking traits, usually between 0.07 and 0.15 (Bugislaus et al., 2005; Cervantes et al., 2020; Ekiz & Kocak, 2005). In horse competitions, heritabilities are neither high when the trait is not the rank, for example: the estimated 0.27 in different studies for annual earnings (Gómez et al., 2010); 0.23 in race time or 0.28 in speed (Ekiz & Kocak, 2005; Ricard & Touvais, 2007); or 0.06 or 0.15 for placing trait (Cervantes, et al., 2020; Ricard & Touvais, 2007). Despite the variance components in the underlying and observed scale can be very different due to the departures from the Gaussian distribution, in rank trait, García‐Ballesteros et al.…”

“…Traditionally, horses participate in several different events, accumulating many records for the animal. Thus, additional random effects are commonly fitted in real genetic models to evaluate horse competitions (Cervantes et al, 2020; Ekiz & Kocak, 2005; García‐Ballesteros et al., 2018; Ricard & Legarra, 2010). Analyses regarding an additional random effect, such as a permanent environmental effect, were carried out to simulate a more realistic scenario, and whether these effects would help accommodate the discrepancy between the stable phenotype and the non‐flat genetic trends.…”

“…Probably the most serious concern is the effect of the competitors' level during the same event when considering ranking traits in genetic evaluation of horses (Cervantes et al., 2020; Gómez et al., 2011). Additional difficulties in establishing a statistical model for a fixed number of participants are: (a) the mean of the qualifying positions is stable over the time leading to no phenotypic trend across time and (b) differences between positions do not directly express the equivalent differences in a supposed underlying variable that would express the performance capacity.…”

“…In summary, by including the event effect in the genetic evaluation any differences would be corrected among the events such as the competition difficulties and the average level of the competitors (Cervantes et al., 2020; Gómez et al., 2011), because participation is not random among events. In this way, the event effect includes the level event difficulty and only depends on the competitors' level.…”

Elo Method and Race Traits: A New Integrated System for Sport Horse Genetic Evaluation

“…Elo rating was estimated only in trotter horses, and the heritability of this trait was 0.27 ± 0.08 [ 18 ]. In another trotter population, a Thurstonian model, which is a competitive evaluation similar to Elo method, provided a lower estimate of 0.09 ± 0.02 [ 12 ]; likewise in Thoroughbred horses, the estimate was 0.047 ± 0.014 [ 33 ] and in other breeds range from 0.07 to 0.17 [ 34 ]. Furthermore, since all the correlations were very high, the trait to choose for maximizing the overall response was the most heritable trait, which was a further advantage.…”