Starch equivalent (SE) values of 345 tropical forages and 37 concentrates derived from digestibility experiments with sheep and cattle formed the data set. Through multiple regression analyses, SE was related to linear, quadratic and cubic terms of the nutrients CP, ether extract (EE), crude fibre (CF), NFE and ash as independent variables. The models were assessed by coefficient of determination ( R 2) and practicality. Under these conditions the following prediction equations were recommended: (a) forages: SE (%) = − 19.48 + 1.4329CP + 2.4494EE + 0.8083NFE; R 2 = 0.46, P < 0.0001, RSD = 6.89%; SE (%) = 134.70 − 1.5703CF − 0.7180NFE − 1.4327ASH; R 2 = 0.45, P < 0.0001, RSD = 6.94%; SE (%) = 65.11 + 0.7427CP − 0.8789CF − 0.7301ASH; R 2 = 0.44, P < 0.0001, RSD = 6.99%; SE (%) = 71.48 − 1.1155CF; R 2 = 0.33, P < 0.0001, RSD = 7.65% (b) concentrates: SE (%) = − 36.46 + 6.1478CF − 0.64 14CF 2 + 0.0158CF 3 + 7.7236NFE −0.1649NFE 2 + 0.0010NFE 3; R 2 = 0.67, P < 0.0001, RSD = 7.76%; SE (%) = 80.15 − 1.2200CF + 1.1005EE; R 2 = 0.42, P < 0.0001, RSD = 10.30%. Crude fibre had a strong influence in the prediction of SE, although it was preferable to predict SE using a combination of nutrients because of a substantial increase in R 2. It was concluded that SE may be estimated using our predictors to within a range of ± 8.8 to 9.8% for tropical forages and to within ± 9.9 to 13.2% for concentrates with a probability of 80%. This level of accuracy was considered high given the diverse nature of the feeds and the wide tropical area for which the predictors may be applied.
Read full abstract