The present study established multiple linear regression models using two ultrasound in vivo measurements (at lumbar and sternal regions, with different real-time ultrasonography machines and probes) and live weight, to predict simultaneously carcass composition and body fat depots of different breeds of sheep and goat. This study is important for the small ruminant industry, considering the feasibility of using the ultrasound methodology in field conditions, as well as an online system of the carcass evaluation. The multiple linear regression models were obtained by selecting the best subset of variables between using the in vivo measurements (raw variables), their second degree and interactions, evaluated in terms of prediction performance using cross-validation “K-folds” and validated by a test group. Overall, high accuracy (adj R²) was obtained from the linear relationship between predicted and experimental values of the group test for each of the nine dependent variables, with values varying between adj R² 0.88 and 0.98.

本研究使用两次体内超声测量（在腰部和胸骨区域，使用不同的实时超声检查仪和探头）和活重建立了多个线性回归模型，以同时预测不同品种的绵羊和绵羊的car体成分和体内脂肪库。山羊。考虑到在野外条件下使用超声方法的可行性以及car体评估的在线系统，这项研究对于小型反刍动物行业而言非常重要。通过在体内使用之间选择最佳变量子集，获得了多个线性回归模型。测量值（原始变量），其第二度和相互作用，使用交叉验证“ K倍”根据预测性能进行评估，并由测试组进行验证。总体而言，从组测试的预测值和实验值对九个因变量的每一个的线性关系中获得了较高的精度（adj R ²），其值在adj R ²介于0.88和0.98之间变化。