A comparison of different models to characterize lactation curves of Friesians in a humid forest zone

Abstract

A study was undertaken of 110 lactations of a herd of Friesians in the humid tropical zone of Ghana to investigate the goodness of fit and other milk production variables estimates by six published lactation curves. These were Yn = a - bn + cn ln (n), (Singh); Yn = n/(a + bn + cn2), (Nelder); Yn = a -bn - ae-cn, (Cobby); Yn = anbe-cn, (Wood); Yn = anb/cosh (cn), (Papajcsik); and Yn = n/ae-bn, (Jenkins); where Yn is the milk yield on day n and a, b and c are the parameters estimated iteratively from the solution to the equations. Season of freshening did not influence peak day, but significantly influenced peak production and 300-day lactation in favour of cows freshening in the major rainy season. Lactation yield was best estimated by Wood model which also gave the highest average R2 values as well as the closest estimate of day of peak production. However, Wood model overestimated peak yields and required the highest number of iterations for convergence. The Jenkins equation imposed a lactation curve that was mostly inconsistent with the data. Except for Jenkins, all models estimated mid to late lactation very well, but differed significantly in their abilities to estimate initial lactation. The relative rankings of the six models by various indexes ranked Singh and Wood models favourably with Nelder as intermediate while the models of Cobby, Papajcsik and Jenkins were ranked poorly.


Ghana Jnl.agric. Sci Vol.32(1), 1999: 31-38