A Comparison of Bimonthly and Quarterly Testing with Monthly Testing for Estimating Dairy Cattle Production 1

One method of effecting economies in product ion testing of da i ry cattle involves increasing the in terval between tests. Several studies have been published concerning the amount and importance of the accuracy lost when this in terval is lengthened. A t least seven investigators have repor ted the accuracy of lactat ion records computed f rom monthly bu t t e r fa t tests and milk weights as compared with an accuracy f rom daily measures (2, 5, 6, 7, 8, 11, 12). Their results indicate an average deviation of the monthly records f rom the daily records of about -+-2 per cent for milk yield and ± 3 per cent for f a t yield. McCar thy and Boyle (7) repor ted tha t 95 out of 100 of the errors they observed in month ly milk records were between + 6.16 and 5.35 per cent. Campbell (2) found tha t 86 out of 100 errors regarding milk yield were less than ± 5 per cent. Seventy-six out of 100 errors in fa t records were within + 5 per cent in his data. Comparisons concerned with the accuracy of comput ing lactation records from weights taken every 2 too. have indicated tha t bimonthly milk yield records vary f rom records computed f rom daily weights on an average of about ± 4 per cent (5, 9). McCar thy and Boyle (7) repor ted that 95 out of 100 observed errors were within ± 10.11 per cent. McDowell (9) found tha t 46 bimonthly records out of 70 var ied f rom the daily records by less than ± 5 per cent. Investigators compar ing bimonthly records with monthly records obtained relat ively similar results (1, 3, 4, 10). Alexander and Yapp (1) found tha t only 49 out of 684 bimonthly fat-corrected milk records var ied more than ± 10 per cent f rom the monthly records. No impor tan t biases were found concerning the accuracy of b imonthly testing. Alexander and Yapp (1) also compared the use of a quar te r ly test with month ly testing. ~Then the three tests were taken on the second, sixth and tenth months of lactation, only 53 out of 684 quar te r ly records erred from the month ly records by more than ± 10 per cent. Two investigations (4, 10) were repor ted which included correlat ion coefficients as a basis for compar ing records computed from vary ing test ing intervals. McCar thy and Boyle (7) pointed out that the correlation coefficient measures the extent to which the types of records fluctuate together, ra ther than measuring the extent to which they are identical. The p r i m a r y objective of the s tudy repor ted in this paper was to estimate