PREDICTION OF GENETIC VALUES OF SIRES FOR GROWTH TRAITS OF CROSSBRED CATTLE USING A MULTIVARIATE ANIMAL-MODEL WITH HETEROGENEOUS VARIANCES

The purpose of this study was to evaluate the effect of adjusting for heterogeneous variances across breed groups on prediction of breeding values (PBV) of selected sires and on breed of sire effects. Data on w eights at birth (BWT), 200 d (WW), and 365 d (YW) of purebred and cros sbred calves from matings of Angus (A), Hereford (H), Polled Hereford, Charolais, Shorthorn, Simmental, Limousin, Maine-Anjou, Chianina, Gel bvieh, Tarentaise, and Salers bulls to A and H cows were used. Calf pe rformance in H and A dams was treated as a different trait. Models com pared included fixed birth year, cow age, and sex classes and crossbre eding effect as a covariate; random direct and maternal genetic and pe rmanent environmental effects were also included, but their variance s tructure was different. Model I assumed homogeneous variances across b reed groups. Model II accounted for heterogeneous variances. Sires wer e ranked based on PBV from each model, and means of PBV of selected si res were calculated based on Model II. Differences between mean PBV we re small for BWT, intermediate for WW, and larger for YW. Differences in PBV of selected sires increased as selection intensity increased, b ut only for WW and YW. Large differences in mean PBV of selected sires between maternal environments (H vs A) were observed for WW and YW fo r various sire breeds. Means of PBV of selected sires based on Model I I exceeded those based on Model I by 6 to 16 kg of YW for various sele ction intensities and maternal environments. Estimates of breed of sir e effects from Model I or II were similar for BWT and WW, but large di fferences were found for YW. Results indicate that some additional eco nomic returns may be gained by commercial producers if sires are chose n across breeds based on predicted genetic values computed with models accounting for heterogeneous variances.