Description of lamb growth using random regression on field data

Abstract

Random regression has been proposed as an accurate method for evaluation of growth data; however, this method has seldom been applied to data from an extensive industry where few records per animal exist. Consequently, field data containing weights of Poll Dorset sheep from 50 to 500 days of age were analysed fitting a random regression model. This model included quadratic, orthogonal polynomials for direct genetic and environmental effects and maternal environmental effects, a linear polynomial for maternal genetic effects and heterogeneous error variances. Direct heritability estimates increased steadily throughout time and were in agreement with literature estimates taken at specific ages. Some estimates for the highest ages with the least records tended to be overestimated, in particular heritability beyond 450 days. Research to solve this problem may require use of a function other than a polynomial. Variances due to maternal genetic effects were low throughout the trajectory. Covariances between weights of sheep for a considerable range of ages can be modelled adequately using random regression.