Prediction of total milk fat of dairy cows: a multi-model approach

摘要

The majority of milk value is detennined by milk protein and fat. Predicting total milk fat has been challenging due to high unexplained variation among and within herds. This study investigated the effect of animal and dietary factors on milk fat yield of dairy cows. Data from 158 studies consisting of 658 treatment means from 2,843 animals (average milk yield = 32.4±6.91 (SD) kg/d) were used. Dry matter intake (DMI) and diet composition were used to calculate fat-free DMI (kg/d), digestible fattyacid intake (g/d), and absorbed amino acids (g/d). A multi-model inference was used to develop a large set of candidate models and the best models were selected using the lowest values of Akaike's Information Criterion corrected for sample size (AfCc) and those models were further evaluated. All models were fitted using a random effect of study and weighted using the square root of the number of animals represented in each treatment. Fat-free DMI, days in milk, absorbed Met, Lys, and He, as well as theintakes of digestible C16:0 and C18:3 were the best variables for predicting milk fat yield (models with lowest AJCc values). The models developed can be used as a practical tool for predicting milk fat yield of dairy cows, while recognizing that additional factors may also affect fat yield.