Comparing Prediction Equations to Estimate Lean Cuts in Swine1

Abstract

Prediction equations were derived to estimate weight of lean cuts (WTLC) in pig carcasses, based on data from 649 pigs, including 163 Duroc gilts, 161 Duroc barrows, 161 Yorkshire gilts and 164 Yorkshire barrows, all produced and processed at Beltsville. Chilled carcasses varied in weight from 68 to 78 kg. The WTLC was for the four standard closely trimmed cuts of picnic and blade Boston shoulders, loins and hams based on whole carcass weights. Various combinations of independent carcass variables (length, backfat thickness, longissimus muscle area and weight), were included in exploratory regression models to compare equations for accuracy of predicting WTLC. Based on R2 and standard errors of estimates, only carcass length of the four variables could be excluded from prediction equations with essentially no loss in accuracy. Separate prediction equations were derived for the different breed, sex and breed-sex subgroups and for all subgroups pooled. Predictions were compared with actual WTLC in the different subgroups. As expected, specific subgroup equations provided accurate estimates of actual WTLC in pigs of that subgroup. However, WTLC estimated by two different pooled prediction equations differed (<.01) from actual WTLC in subgroups of Duroc barrows, Yorkshire gilts, all Durocs and all Yorkshires. In separate analyses, the effects of longissimus muscle area and carcass weight differed (<.01) between barrows and gilts, but no practical advantage was noted for including a quadratic effect in the prediction equation. It was recommended that the problem of biased estimates be addressed by further research to derive prediction equations to estimate lean cuts or lean tissue content for use in more specific populations, primarily with barrows and gilts.