Modelling fat and protein concentration curves for Irish dairy cows

The objective of this study was to acquire a well-fitting, single-equation model that would represent the fat and protein concentration curves of milk from Irish dairy cows. The dataset consisted of 16,086 records from both spring and autumn calving cows from both experimental and commercial herds. Many models cited in the literature to represent milk yield were examined for their suitability to model constituent curves. Models were tested for goodness-of-fit, adherence to the assumptions of regression analysis, and their ability to predict total fat and protein concentration for an entire lactation. Wilmink's model best satisfied these criteria. It had the best Mean Square Prediction Error (goodness-of-fit) value, it satisfied the assumptions of regression analysis (multicollinearity, heteroskedasticity, autocorrelation and normality of distribution), and it predicted the actual concentration of the constituents to within 0.01 percentage point.