摘要：Different AT milk recording schemes, with purpose of detection of scheme with the highest lactation milk yields prediction accuracy as well as with simplest practice application, were compared in this study. Analysed data were collected from November 2004 till November 2006 on 15 family farms in Croatia according to A4 milk recording scheme by the field assistants of the Croatian Livestock Centre. Measuring at test-day included measurement of milk yield and taking one sample for milk composition analysis at each milking (evening and morning). During research, total of 769 cows were included in research. More than 50 % of all cows were Holstein breed (58.9 %), while the rest of the cows belonged to the Simmental breed. For prediction of daily milk yields two methods were used: M I. = linear regression of daily on partial values with taking into account the interval between successive milkings; M II. = doubling of partial values. The Test Interval Method (TIM) was used for prediction of lactation values. As evaluation criteria for comparison of schemes, correlation and difference between estimated and true milk yields were used. Scheme S II., that is use of first morning, then evening test-day record, in case of doubling the partial yields, showed as scheme with the highest prediction accuracy. Regardless of milk recording scheme, use of method M I. in daily milk yields prediction, resulted in high accuracy and low milk yields, milk fat and protein bias prediction in standard lactation. Research results implicate the use of method M I. in daily milk yields prediction, as well as possibility of different AT milk recording schemes combination in practice.
