A comparison of alternative methods for predicting the carcass composition of crossbred lambs of different breeds and crosses.

Carcass data for a total of 2808 lambs from three breed trials were used to evaluate a series of linear measurements, visual scores and the proportions of tissues in joints as predictors of carcass composition. The trials involved crossbred lambs out of different dam types by sires of the main British meat breeds and by Ile de France, Oldenburg and Texel sires. Each trial was carried out over a number of years and involved approximately equal numbers of female and castrated male lambs. Potential predictors were examined in terms of the precision of prediction and in terms of the robustness of prediction equations to differences in sire breed and sex. The overall standard deviation (s.d.) of carcass lean percentage averaged over trials was 4·3. Combinations of simple measurements, including visual fat scores, percentage perinephric and retroperitoneal fat in carcass and M. longissimus dimensions achieved a residual s.d. of 2·5 for carcass lean percentage. The application of an overall prediction equation to individual sire breed means resulted in considerable bias (predicted-actual lean percentage): the mean square deviation was 0·75. In comparison, the sex differences were relatively unimportant. The precision of sample joints was examined in relation to their cost of dissection. The best end neck and shoulder joints offered a high level in relation to cost: typical residual s.d. were 1·5 for carcass lean percentage. Joints and combination of joints with high predictive precision tended to have equations that were robust to differences between sire breeds.