By George Wiggans, Paul VanRaden, and Lillian Bacheller
For some records, the herd averages used by best prediction in estimating yields between test days or after the last recorded day were changed. Previously, the rolling herd average reported on the last test day of the lactation was used, and if no herd average was reported, a breed average was substituted. A new procedure introduced in August increases the frequency that herd averages are used instead of breed averages. Some herds only test for SCS every other month, and the herd average is missing for the untested month. Also, some herd averages were not matched correctly if the herd average was stored with a different milking frequency than that reported for the cow. Now we search backward, separately by trait, through all the test days for a lactation to find the latest matching herd average. This change was applied to all calvings since January 1997. Also, updated breed averages were obtained from the 2005 base cows (cows born in 2000), and these are used for records initiated after January 2000 if the herd average is missing.