To establish the measures that best define the quality of the carcass of swine, data of lean meat yield (MY), fat yield (FY), dressing percentage (DP), carcass length for the Brazilian Method of Carcass Classification (CLBM), carcass length for the American Method of Carcass Classification (CLAM), average backfat thickness (ABT), backfat thickness 4 cm from the dorsal mid-line (P1), backfat thickness 6.5 cm from the dorsal mid-line (P2), backfat thickness 8 cm from the dorsal mid-line (P3), loin eye area (LEA), lean:fat ratio (LFR) and ham yield (HY) were used, obtained in 664 Large White swine. Initially, in order to make clear the importance and possible relations structures among variables, the cluster analysis was used. Afterwards, the multiple regression linear model was used, using the SAS package. Firstly, it was estimated the partial correlation (Type II) of each trait to evaluate the importance of the quality variables of the carcass. Secondly, a procedure of adjusting equation selection named Backward was used. It was verified that, except for DP, all the measures of carcass classification evaluated presented correlation with MY and FY. The measures ABT, P1, P2, P3 and LFT were inversely correlated with MY and directly correlated with FY. The opposite was observed for the measures CLBM, CLAM, LEA and HY. ABT, AOL, RCG and RP are just enough to evaluate the quality of the carcass in terms of lean meat yield, and, in terms of FY, besides these measured previous it's necessary the evaluation of P1. The measures of loin eye area and average backfat thickness, as proposed by MBCC, shown as being what best evaluates the amount of meat and fat in the carcass.
correlation; cluster analysis; meat and fat yield
