Random regression models for the prediction of days to finish in beef cattle

The idea of reducing the number of days required for livestock to reach their desired endpoint is not new, with its economic importance first discussed in 1957. Given this economic relevance, genetic evaluation research for reducing these required days has received very little attention throughout the pertinent literature with the exception of the swine industry. Many different production systems exist in today's beef industry, and a single prediction for the required number of days to reach a single finish endpoint does not lend itself well to this diversity. Because of this point, random regression models are an attractive alternative to more traditional multiple trait or repeated measures best linear unbiased prediction models in the calculation of days to finish.;Random regression models estimate regression lines for each animal in the pedigree, thereby resulting in the ability to calculate estimated breeding values (EBV) for any age or any number of days on feed. This inherent property allows beef producers to calculate days to finish EBV for finish endpoints that fit individual production scenarios.;The objective of this study was to develop a series of models using random regression techniques for the genetic prediction of the required number of days to reach the finish endpoints of weight, ultrasound back fat and ultrasound rib eye area. This study performed some basic tasks of describing data and the behavior of random regression models used for the prediction of days to finish.;Genetic predictions for the traits days to weight (DTW), days to ultrasound back fat (DTUBF) and days to ultrasound ribeye area (DTUREA) were prototyped using data obtained from the Agriculture and Agri-Food Canada Research Centre, Lethbridge, Alberta. This data consisted of pedigree, weight, ultrasound back fat and ultrasound ribeye area observations on 1,324 cattle spanning the years 1999 -- 2007. Individual animals averaged 5.77 weight observations with weights and ages ranging from 293 kg to 863 kg and 276 to 519 days, respectively. For the ultrasound traits individual animals averaged 5.57 observations. Ultrasound back fat observations ranged from 1.53 mm to 30.47 mm and ultrasound rib eye area observations ranged from 36.77 cm2 to 129.54 cm 2.;Fixed effects included in the model were determined through a series of regressions to identify those accounting for a significant amount of variation in the age response variable. Results showed for the trait DTW the effects of year, pen and breed type should be included, and due to the confounding of year and breed type, all three were included in the contemporary group definition. Similar results were obtained for both DTUREA and DTUBF. Year of measure, pen and breed were included in the contemporary group definition for both traits. Using these three effects to form contemporary groups resulted in average contemporary group sizes of 21.50 and 21.45 for the days to weight and days to ultrasound traits, respectively. All three models, contained the effects of contemporary group and a fixed regression of age on weight / ultrasound back fat / ultrasound rib eye area to account for the overall mean relationship between age and each of the three finish traits.;Random regression models were built for each of the days to finish traits. Model building exercises for the three traits consisted of conducting likelihood ratio tests to determine the order of the random regression polynomial. For DTW, a linear random regression polynomial was sufficient in describing the genetic variation in days. Depending on how residual variance was modeled, heritability estimates varied. When observations were classified into four distinct residual variance sub-groups, heritability estimates for DTW ranged from 0.56 for the number of days to reach 293 kg all the way to 0.93 for the number of days to reach 863 kg. If residual variance was modeled using a linear random regression, heritability estimates for DTW were more conservative ranging from 0.53 for the number of days to reach 293 kg to 0.76 for the number of days to reach 863 kg.;The significant random regression order for the ultrasound traits was dependent on how the residual variance was modeled. For DTUREA, when residual variance was modeled using four distinct sub-groups, a quartic random polynomial was needed to model the genetic variation in days. When a linear random regression was applied to the residuals, a linear polynomial was all that was needed. The quartic polynomial tended to artificially inflate heritability estimates in the extremes of the data distribution for DTUREA ranging from 0.81 (36.77 cm2) then dropping to 0.15 around 110 cm2 and jumping back up to 0.91 at 129.54 cm2. Heritability estimates obtained from the linear random regression using linear residual random regression were much more sensible, ranging from 0.53 at 36.77 cm2 to 0.49 at 129.54 cm2.;For the trait DTUBF, when residual variance was modeled using four distinct sub-groups, a quadratic random polynomial was all that was needed to describe the genetic variation in days. Similar to DTW and DTUREA, the linear residual random regression model only needed a linear polynomial. Heritability estimates for DTUBF from the linear random regression model using linear residual random regression ranged from 0.54 at 1.53 mm of ultrasound back fat to 0.35 at 30.47 mm of back fat. Heritability estimates from the four residual sub-groups model became much more variable ranging from 0.58 at 1.53 mm of back fat down to 0.08 at 26 mm of back fat then jumping back up to 0.54 at 30.47 mm of back fat.;For all three traits, modeling the residual variance using a linear random regression seemed to be the most ideal, as it required the lowest order polynomial for describing the genetic variation in days. The linear residual random also yielded the most realistic heritability estimates for each of the endpoints. Heritability estimates obtained in this study show the days to finish traits are moderately to highly heritable, depending on endpoint. As such, sufficient genetic variation exists to make fairly rapid progress in reducing the number of days to reach finish endpoints, giving producers tools to increase the profitability of their operations.