[en] Reducing the frequency of milk recording and the number of recorded samples per test-day could be a solution in order to reduce costs of milk recording. However, this solution leads to decrease also the accuracy of predicting daily yield. According to the literature, several authors have already worked on this problematic. Unfortunately, some effects used in previous studies are not often available or reliable in used databases. This study was aimed to enlarge these investigations to milk fatty acids (FA) production: saturated FA, mono-unsaturated FA, unsaturated FA, medium-chain FA, and long chain FA and to propose a simple, robust and practical method for estimating accurate daily major FA yield from single milking. To do this, five dairy cows were followed between January 2007 and December 2010. FA contents were predicted by mid-infrared spectrometry. The final database contained 1,440 records. The first step was to ensure that used effects were available in most used databases. According to the availability of data, height models were tested to estimate daily yields from both morning and evening milking. These models were compared on the basis of the coefficient of determination values between estimated and observed daily yields and the mean square error. The proposed models included progressively several effects such as the milk yield, the fat and protein content, some classes of stage in lactation, of month of test or of month of calving. As expected, R² values were higher when these effects are introduced in the model and were comprised between 0.87 and 0.88 when daily yield were estimated from morning milking, and from 0.75 and 0.86 when daily yield were estimated from evening milking. It was concluded that the introduction of these effects did highly improve the daily predictability of all trait yield and can partially replace the milking interval effect. It was also observed that daily yields estimated from evening milkings are less accurate than those estimated from morning milkings. Finally, the applied model will depend on the availability of the data and to the convenience of the applied model to the studied population.
Keywords: Milk recording, Fatty acids, prediction
