Evaluation of the accuracy of DHIA predicted milk using actual milk shipped

Abstract

The purpose of this study was to compare DHIA predicted monthly milk values to actual monthly milk shipped values and identify factors affecting the over or under estimation of actual values. Data on 454 Texas herds from January 1991 through May 1995 (53 months) was obtained from DRPC@Raleigh. The Milk Market Administrator Federal Order, Number 126, Carrollton, Texas contributed data on 328 herds for the same period. After combining and reconstructing the data, 286 herds remained in the study. A statistical model which included season, type of test, bST, herd(test) and three interactions was used to determine the percent DHIA estimated monthly milk values over or under estimate actual monthly milk values shipped. reason was found to be significant(p<.Ol) in the model as was type of test (p<.05). A bST variable was introduced into the model to test for an effect due to the commercial introduction of bST. The variable was based on date, where the time period before February 1994 was considered pre and the period after February 1994 was classified as post. The bST variable showed significance in the model (p<.Ol), as did the bST X type of test interaction (p<.01). In order to make comparisons within the data, contrasts were used to group the levels of variables for comparisons of interest. Types of tests were grouped for comparison in three ways: Official versus Unofficial, DHI versus DHK and 2X versus APCS versus APT. Official records were compared to Unofficial records and showed a significant difference (p<.O I) with Unofficial records being closer to actual values than Official records. DHI types of tests when contrasted against DHIR values, showed no significant difference. Comparisons between APCS tests and APT tests showed significance (p<.05) as did the contrast of APCS versus 2X types of tests (p<.05). 2X tests did not show a significant difference from APT test, however.