Estimation of segregation variance for birth weight in beef cattle

Abstract

Genetic evaluation using multibreed covariance theory requires estimating the segregation variance. The segregation variance is the amount by which the additive variance in the F2 exceeds that in F1. The goal of this research was to obtain REML estimates of the additive variances plus segregation variance, assuming equal environmental variances for all genetic groups. The data were originated in two experimental herds of beef cattle from New Zealand (NZ) and Argentina (AR). Records were birth weights of 4082 Angus-Hereford (NZ) and 6963 Nellore-Hereford (AR) cross calves, including purebreds, F1, backcrosses, F2, and advanced generations (F3, F4, F5). Variance components were estimated using an additive animal model by REML, with a first-derivative algorithm. The asymptotic standard errors of the REML estimates were calculated using the inverse of the information matrix. After 400 iterations, estimates of the additive variances (in kg2) were 7.77±0.91 (Angus) and 10.02±1.11 (Hereford), and estimate of the segregation variance was 1.14±0.85, in NZ data. Whereas in AR data, estimates of the additive variances were 6.59±0.71 (Nellore) and 8.97±0.75 (Hereford), and estimate of the segregation variance was 1.48±0.74. The error variances were estimated to be 7.92±0.06 in NZ and 6.86±0.06 in AR. Asymptotic tests (Likelihood Ratio and Lagrange Multiplier) of the hypothesis of null segregation variance suggested that this was not the case in these data.