Date of Award
Doctor of Philosophy
Don O. Richardson
J.T. Miles, M.J. Montgomery, L.M. Josephson, R.R. Shrode
A sample of 471 lactation records from four dairy herds of The University of Tennessee were utilized in this study. The experiment was designed to investigate the possibilities of lengthening the DHIA testing interval. The study was limited to 305-day milk yields. Three sampling intervals: (1) 30 days, (2) 60 days and (3) 90 days and three computing methods: (1) unadjusted, (2) first-test-adjusted and (3) first-and-last- test-adjusted were used to estimate the lactation milk yields. Each lactation record provided 30 monthly, 60 bi-monthly and 90 tri-monthly unadjusted estimates. In all, 254,340 lactation estimates (adjusted and unadjusted) were computed.
The deviation of estimate from the actual yield was measured as percent of actual yield. Statistical analysis indicated that herd, age, interval and computational methods produced variation in the percent deviations of estimates. Interactions of these factors also caused variation in percent deviations in most cases. In all the intervals investigated, the average error and the average percent absolute deviation was greater for higher production levels. The average errors of monthly unadjusted, first-test-adjusted and first- and-last-test-adjusted estimates were 0.63, -0.02 and -0.13 percent of actual yields, respectively. The corresponding values for bi-monthly estimates were 0.84, 0.21 and -0.24 percent, respectively. This indicated that in monthly and bi-monthly intervals, the last test correction in addition to the first test correction was not of any additional value. The conclusion was also supported by similar trends in the average percent absolute deviations, frequency of estimates within a given percent of actual yield and standard deviations of percent deviations. The average errors of tri-monthly (1) unadjusted, (2) first-test-adjusted and (3) first-and-last-test-adjusted estimates were 0.84, 0.87 and -0.14 percent of actual yields, respectively. The corresponding values for absolute deviation were 1.57, 1.75 and 0.85 percent of actual yields. The first test correction by itself was of no value in improving the tri-monthly estimates, but correction of first and last test was essential.
The average errors of bi-monthly first-test-adjusted and trimonthly first- and-last-test-adjusted estimates were closer to zero than that of monthly adjusted estimates. A study of estimates in three parts of the lactation indicated that as the sampling interval was elongated, the error from the last one-third part of the lactation became more important. The effect of correction of tests was not the same in all herd-age (production) levels. A study of average deviations of estimates by day of first test indicated that delaying testing up to about the tenth day of lactation would decrease the average error as well as variation of percent deviation of tri-monthly estimates.
Hence, if more appropriate factors were periodically evolved on a region-breed-age basis for correcting the estimates, and if the first tests were delayed to about the tenth day of lactation, the tri-monthly first- and-last-test-adjusted estimates could be an alternative to the monthly unadjusted estimates. The savings on tests and computation thus achieved could be passed on to the farmer which might result in increased participation of the farmers in DHIA testing.
Rai, Amai Vishwanatha, "A comparison of different sampling intervals and computational methods for estimating lactation yields. " PhD diss., University of Tennessee, 1971.