Fitness of Lactation Curve Functions to Daily and Monthly Test-Day Milk Data in an Ethiopian Multi-Breed Dairy Cattle Population

Abstract

The objectives of this study were to identify the lactation curve function that had the best fi t to daily and monthly test-day milk data and to evaluate the factors affecting parameters of the best fi t lactation curve function for an Ethiopian dairy cattle population. An incomplete gamma (IG), a modifi ed incomplete gamma (MIG; b = 1) and an inverse polynomial (IP) function were compared using 6,717 lactation milk records of 2,064 cows from the Bako, Holetta and Debre Zeit Research Centers, Ethiopia. Breed groups were Horro (H), Boran (B), B × Friesian, H × Friesian, B × Simmental, H × Simmental, B × Jersey and H × Jersey. The MIG and IG were log-transformed to linear form before fi tting. The functions were compared based on the least squares means (LSM) of R2 (LSM R2) and adjusted R2 values and on the accuracy of lactation milk yield prediction. The statistical model included herd-year-season of calving, parity, data type, breed group, lactation curve function, and the interactions of data type × function and breed × data type × function as fi xed effects, and the residual as a random effect. The MIG, IP and IG functions ranked from the best to the worst fi t based on LSM R2 and adjusted R2. The LSM R2 and adjusted R2 were signifi cantly (P < 0.001) different among all classes of fi xed effects considered in the model. The LSM R2 and adjusted R2 for the MIG function were 0.90 and 0.89, respectively. All functions fi tted to monthly test-day better than to daily milk data. The MIG function had the best fi t (P < 0.001) to daily milk data, but both the MIG and IP functions had a similar fi t to monthly test-day milk data based on the LSM of adjusted R2. The ln(a) and c from the MIG function with daily and monthly test-day milk data, and the A0, A1 and A2 from the IP function with monthly test-day data were different among breed groups, parities and herd-year-season classes (at least P < 0.05). The MIG function predicted the lactation milk yield from the monthly test-day milk with the lowest prediction error (P < 0.001) compared to the IP and IG functions. Thus, the MIG function could be recommended to model lactation milk data from monthly test-day milk in the studied dairy cattle population.