A simple regression was as effective as multiple regressions for 30 carcasses (Group 1), but for each equation there was a significant selection-line effect on the intercept. When the usual direct regression coefficient was replaced by an inverse or a geometric co- efficient, the selection line effect was not significant.
The 3 prediction equations were evaluated by comparing predicted side subcutaneous fat percent with actual values for a separate lot of 48 ram carcasses (Group 2). The correlation between predicted and actual values was high, but the selection-line effect was significant for all 3 equations. These results serve to emphasis that even within a breed and sex, and for relations with high correlations, there may be problems with the stability of the prediction equations. This could introduce a bias into conclusions.
Keywords: NZSAPAB; Southdown; rack cut; subcutaneous fat; genetic line; sample cut