A Computationally Feasible Test Day Model for Genetic Evaluation of Yield Traits in the United States

Abstract

A multitrait analysis of test day yields is proposed that includes 60 traits [3 yield traits (milk, fat, and protein), 2 parity groups (first and later) per yield trait, and 10 stages of lactation per parity]. To reduce the computations needed for the 60 traits, test day effects are estimated within the herd before analysis across herds, the rank of the genetic (co)variance matrix is reduced, a canonical transformation is used with missing values replaced by their expectations, and a repeatability model is applied to allow inclusion of parities after second. Historical 305-d records are included through their correlations with test day effects. Possible benefits from this model include 1) more accurate estimation of environmental effects from including the influence of particular days of recording, 2) optimal use of information from all test days (especially for lactations with long intervals from calving to first test or between tests), 3) improved accuracy of evaluations for component yields through contributions from information for milk yield, and 4) greater stability of bull evaluations from accounting for genetic differences among daughters in the shape of lactation curve and maturity rate.