Comparison and evaluation of mathematical lactation curve functions of Iranian primiparous Holsteins

Abstract

Describing lactation in mammals using a lactation curve aims to provide a concise summary of the pattern of milk yield and valuable information about the biological and economic efficiency of the animal or herd under consideration. A total of 106 581 monthly test-day milk records collected from 12 677 Tehran Province primiparous Holstein cows from 151 herds, were used in this study. Using the General Linear Model procedure in SAS, the effect of herd, calving year, age at calving, season of production, age and days in milk were found to be significant on daily milk yield. The suitability of seven mathematical models (with three, four and five parameters) for describing the 305-day milk yield lactation curve of Holstein cows, were examined in this study. Comparisons of the models were made based on the coefficient of determination, root mean square error, Durbin Watson coefficient and sum of daily deviations, by using nonlinear (NLIN), regression (REG) and autoregression (AUTOREG) procedures in SAS. The best three, four and five parameter functions with respect to these criteria were the Incomplete Gama, Dijkstra and Grossman functions. With regards to the Wilmink model, the best results obtained were from models with the constant of 0.05 and 0.065. The Wood function was selected as the best model for prediction of daily milk yield, due to parameters in the function and less computational limitation.

Keywords: Daily milk yield, goodness of fit, lactation, model comparison