PREDICTION OF ENERGY BALANCE USING PRINCIPAL COMPONENT REGRESSION OF MILK COMPOSITION MEASURES
Keywords:
energy balance, milk composition, Friesian –Bunaji crossesAbstract
The purpose of this study was to investigate whether there is a stable relationship between energy balance (EB) and milk composition measures, and, how accurate could EB be predicted using group mean data. Energy balance was evaluated using data in which milk composition was measured at each milking from thirteen (13) primiparous and 47 multiparous (F1) Friesian x Bunaji cows. The relationship between EB and milk measures was quantified by Principal Component Regression (PCR). The initial model included 17 combinations of milk composition measures allowing a range of nonlinear effects. These combinations were as follows: DMY, TMY, MFC, MPC, MLC, MFY, MPY, MLY, FPR, FLR, PLR, together with 10 milk composition change variables “diff()”. These are the current minus the previous values of the milk measures in question {diff(my), diff(MFC), diff(MPC), diff(MLC), diff(mFy), dff(mPy), diff(mLy), diff(FPR), diff(FLR), diff(PLR)}.Using group mean data, a very high proportion of the variability (Adjusted R2) in EB (over 99%)was explained by the PCR models. The prediction errors (RMSEP) for the PCR models were low, ranging from 7.54 to 18.3MJ/d. Therefore, using the PCR procedure that based only on milk composition measures, EB was predicted with high degree of reliability both within and across lactation stages. The potentials indicators of EB were the rate of change composite variable. Based on this observations, it is concluded that the mean EB of Friesian x Bunaji cows can be predicted throughout lactation using a common equation base on milk yield, milk fat and milk protein measures.