Evaluation of Alternative Methods to Develop Prediction Equations to Predict Carcass Fat-Free Lean Mass1

Carcass measurements of 203 pigs representing 7 genetic populations and 4 target weights (100, 114, 128, and 152 kg) were used to evaluate alternative equations to predict carcass fat tissue-free lean mass. Independent variables in the equations included carcass weight and measures of backfat depth and longissimus area. Alternative sets of equations were calculated to evaluate the inclusion of a random effect of week of processing and of the quadratic and cross-product independent variables. In addition, equations were developed from data including the overall data set as well as from the 3 lightest and 3 heaviest target weights. Overall, the addition of the random effect of week did not change the residual standard deviations of the prediction equations. For 4 equations, the inclusion of the random effect of week reduced the absolute value of the regression coefficients. The best equations developed from either the light or heavy pig data sets had less total residual sums of squares and less residual sums of squares accounted for by target BW group than did the best equations developed from the overall data set. In 2 of the 4 cases, the best linear equation developed from the light or heavy data sets had less residual sums of squares than the best equation based on the overall data set.