Prediction of the useful energy in tropical feeds from proximate composition and in vivo derived energetic contents .1. Metabolisable energy

Multiple regression models were used to predict ME from the proximate composition of 345 tropical forages and 37 concentrates. Metabolisable energy was derived from published digestibility experiments with sheep and cattle. Linear, quadratic and cubic terms of the proximate components were used in the regression analyses. Predictors were derived under the conditions of maximum coefficient of determination (R-2) and practicality. On the basis of both criteria, the following predictors were selected: (a) forages: ME (MJ kg(-1) DM) = 8.11 + 0.1341CP - 0.1065ASH; R-2 = 0.21, P < 0.0001; ME (MJ kg(-1) DM) = 20.27 - 0.1431CF - 0.1110NFE - 0.2200ASH; R-2 = 0.25, P < 0.0001; (b) concentrates: ME (MJ kg-l DM) = 5.34 - 0.1365CF + 0.6926NFE - 0.0152NFE(2) + 0.0001NFE(3); R-2 = 0.45, P < 0.0001; ME (MJ kg(-1) DM) = -4.80 + 0.6004CF - 0.0640CF(2) + 0.0015CF(3) + 1.1572NFE - 0.0236NFE(2) + 0.00014NFE(3); R-2 = 0.54, P < 0.0001. Correlation coefficients were of the order of 0.5 when predicted ME values were validated against known ME values for different forages in eastern and southern Africa. It was concluded that the predictors developed were simple and practical and could be used with an 80% accuracy to predict the metabolisable energy value of tropical forages and concentrates allowing for 1.4 MJ ME deviation from the mean. (C) 1997 Elsevier Science B.V.