AJUSTE DE MODELOS PARA PREDECIR EL RENDIMIENTO Y CARACTERÍSTICAS DE LA CANAL EN BASE A VARIABLES PRODUCTIVAS ANTEMORTEM DE OVINOS EN ENGORDA CON ALIMENTACIÓN INTENSIVA

Abstract

<p><strong>Background: </strong>Instrumental evaluation of carcass characteristics, meat quality and sheep performance, require specialized equipment, therefore it is necessary to have available technological and economic resources, which sometimes result expensive throughout the meat chain production value of sheep. Prediction of sheep carcass characteristics based on mathematical models, is a good, economic, confident and repeatable method. <strong>Objective: </strong>To adjust, through two methods of estimation, prediction equations of <em>postmortem</em> variables by means of <em>antemortem</em> productive variables of intensive fattening sheep slaughtered in the valley of Toluca, Estado de México. <strong>Methodology:</strong> A total of 175 records of fattening sheep, slaughtered in small slaughterhouse of barbecue cookers in Capulhuac Municipality of the Estado de México were used. There were used 8 <em>antemortem</em> variables, in order to estimate prediction equations of 18 variables associated with performance, morphometry, muscular conformation, and grade of sheep carcass greasy. Carcasses were classified according with their similarity and grouped in principal components (PC), then were carried out multiple linear regression (MLR) analysis over original variables and factorial loads with extraction methods of principal components (PC). <strong>Results: </strong>The adjusted equations with MLR, showed a <em>R</em><sup>2</sup> ≥ 0.42 for HCW (hot carcass weight), CCW (cold carcass weight), LP (leg perimeter), RW (rump width), TD (thorax depth), and CI (compactness index). The assumptions of MLR were verified and the statistics Tol, VIF, DFBETAS and DFFITS demonstrated multicollinearity between variables. For the regression analysis, the principal components (RPC), were obtained three PC that explained 82.78% of the σ<sup>2 </sup>(variance), and the adjustment of MLR over factorial loads obtained equations for HCW, CCW, PL, RW and CI with <em>R</em><sup>2</sup> ≥ 0.37, up to 0.73. It should be noted the importance of the adjusted equation for CCW because of its relation with carcass price and its weight as a predictor variable of primary and commercial cuts. <strong>Implications</strong>: It is useful and necessary that the adjustment of prediction equations for performance variables in animal science, can be accompanied with results of their respective tests of the model assumptions of multiple regression analysis. Our findings, confirm the need to carefully examine the adjustment of prediction equations with the aim of estimating equations with less bias and higher confidence. <strong>Conclusions: </strong>The multiple regression analysis over original variables and vectors of principal components determined prediction equations with different grades of adjustment for performance variables (HCW, CCW, CI) and carcass quality (LP, RW, TD). In the adjusted equations over original variables, the betas with higher prediction power were for slaughter weight, initial live weight and final live weight. While for the adjustment of prediction equations with factorial loads of PC, the betas with higher power of prediction were for PC1 y PC2, characterized by having higher factorial loads. The values of multicollinearity and autocorrelation bias, determination coefficient and explained variance, showed that practical application of these prediction equations allowed to a real approximation to estimation of <em>postmortem</em> variables; however, these values should be taken considering their reliability.<strong></strong></p>