Comparison of Repeatable and Random Regression Models for Genetic Parameter Estimation on Thoroughbreds

Abstract

Introduction - The combination of multiple factors (track, year, ages, etc.) is effective in achieving the maximum level of race performance in Thoroughbreds. Aim - The aim of this study is to estimate and compare genetic parameters on the number of race success characteristics in Thoroughbreds with random regression models (RRM) and repeatable animal models (RAM) with a different number of repetitions. It was also aimed to investigate which number of observation points would be sufficient for genetic parameter estimation for Thoroughbred. Materials and methods - As data, 111312 test day race completion time (sec) records of 13625 Thoroughbreds raced taken from the Jockey Club of Turkey between 2005 and 2016 were used. Competition performances were compared with different measurements using the same repeatability model. Variance components of Thoroughbreds were obtained by using RRM and RAM using DFREML and WOMBAT package, respectively. Results and discussion - When AIC and BIC values were examined, it was observed that the values in RRM were lower than RAM method for ten races. According to Akaike Weights results, while the fifth race shows 35.56% better fit than the fourth race in the model. The AW values of other number of races showed less superiority; thus, 5th and 6th races can be preferred over its competitors in terms of Kullback-Leibler discrepancy. Conclusion - Our results showed that number of race of five were sufficient to estimate genetic parameters for Thoroughbred horse. Also, RRM method can be preferred compared to RAM method.