Treatment of Deviant Lactation Records in Sire Summaries

Open Access
Publisher Website
Google Scholar
Abstract
Effects of procedures applied to widely deviant records on a mass basis to improve predictive ability of early Sire Summaries were investigated. Data were first lactation daughter records for 200 Holstein bulls with at least 500 daughters. Five subsets of the first 10, 30, 50, 70, or 90 daughters to calve for each bull were created to simulate data ordinarily available for the first few Sire Summaries in a bull''s life. Three empirical procedures were developed based on SD appropriate for herd yield: 1) differences between daughter Modified Contemporary Deviation and sire ancestor merit were multiplied by probability function values (regression procedure), 2) length weights for calculating daughter Modified Contemporary Deviation were multiplied by probability function values (weighting procedure), and 3) records were excluded based on number of SD between daughter Modified Contemporary Deviation and sire ancestor merit (elimination procedure). Experimental site evaluations were calculated and compared with July 1984 USDA-DHIA Sire Summaries. Maximum improvements of 1 to 3% in predicting genetic merit were found for all three procedures for evaluations based on smaller numbers of daughters. If these small increases in accuracy are repeatable, they would justify added computational expense necessary to implement adjustment procedures for a national data set.