Estimation of body composition in tropical sheep raised under seasonal feed supply conditions: Prediction models

Prediction models were developed from isotope dilution space (D2O) and live animal measurements (heart girth, height at withers, body length and tail volume measurements) to estimate chemical body components of indigenous tropical fat-tailed sheep breeds in vivo. A STEPWISE multiple regression procedure of SAS was used to assess the predictive power of combinations of variables and models which minimise the predicted residual sum of squares. With regard to the accuracy and robustness of prediction, models containing body weight as the only predictor variable resulted in less accurate estimates of body components, especially that of body fat and energy contents. However, the use of isotope dilution space (as an index of Total Body Water) along with body weight measurements showed significant improvements in R² and accuracy of prediction equations. Testing the predictive ability of models containing live animal measures only, the result obtained showed that, despite a small reduction in accuracy, indices of live animal measures gave comparable estimation of body components with models containing isotope dilution space. Therefore, considering the cost of D2O and its applicability in the field, indications are that the use of models containing indices of live animal measurements only (in various combinations) is promising for field applications, to provide longitudinal measures of change in body composition of tropical fat-tailed sheep