A comparison of different predictors of the lean content of pig carcasses 2. Predictors for use in population studies and experiments

Abstract

Linear measurements and sample joint dissections were examined as predictors of carcass lean percentage using data from 1060 carcasses comprising 108 subgroups (nine genotypes × two sexes × two feeding regimes × three slaughter weights). The potential predictors were examined in terms of precision, cost and the stability of their regression equations.Probed fat thickness 6·5 cm from the dorsal mid-line at the last rib (P₂) was the individual measurement giving the most precise prediction (residual s.d. pooled within subgroups = 2·21 percentage units; r = −0·72). The square root transformation (P₂^½) gave stable regression slopes over levels of all four factors, though there were important differences in intercepts between genotypes, sexes and slaughter weights.The precision of prediction from the lean percentage of sample joints was generally improved by including P₂^½ with each in multiple regression. Used in this way with current Meat and Livestock Commission costings, the ham joint offered the best compromise between cost and precision (pooled residual s.d. = 1·25 percentage units; R = 0·92). Regression slopes on each joint were stable for all factors except for those on the rump back and rib back joints which varied by feeding regime. Intercepts varied by genotypes for all joints except the ham, while intercept differences for feeding regimes were small or non-significant.The results confirm the value of the ‘sub-sampling with regression’ procedure for predicting carcass composition.