Accounting for subpopulations in prediction of the proportion of lean meat of pig carcasses

Abstract In a dissection trial in The Netherlands two subpopulations were distinguished: gilts and castrated males. The sampling scheme, which emphasizes extreme values for the proportion of lean meat in the carcass, was followed for the two sexes separately, to ensure sufficient accuracy for a comparison between them. Significant differences between the prediction formulae for the lean meat proportion for the two sexes were found. Since it is not possible to use separate prediction formulae for the sexes in Dutch slaughterhouses, the formulae had to be combined into one overall prediction formula. In this paper it is shown how the separate prediction formulae for the sexes may be combined, utilizing additional data, not involving dissection, which were easily collected on the slaughterline, at little extra cost. The method can be extended to cover any number of subpopulations. Two objectives can be achieved at the same time: subpopulations may be compared accurately on the basis of a stratified sample and from the results of the comparison an efficient, unbiased, overall prediction formula may be distilled.