Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: March 23, 2004
Publication Date: July 25, 2004
Citation: Wiggans, G.R., VanRaden, P.M. 2004. Accounting for differences in rate of maturity in yield evaluations [abstract]. Journal of Dairy Science. 87(Suppl. 1):412. Technical Abstract: Genetic evaluations that include repeated records usually assume genetic merit is the same across all lactations. Repeatability models are simpler to set up and solve than models with separate genetic effects for each lactation, but may be less precise if animals differ in rate of maturity. Several countries now model the first three lactations as separate traits in published evaluations for yield traits. An alternative is to include a regression on maturity in traditional repeatability models. This allows data from five or even all lactations to be modeled by two genetic effects using single-trait rather than multi-trait programming. Maturity effects were scaled to measure the difference between first and second lactation yield, with smaller differences among later parities. The model was tested on US Jersey data of over 3 million lactation records. Maturity regression coefficients for lactations 1 to 5 were set to [-0.7 0.3 0.4 0.4 0.4] based on preliminary analysis. Parent averages calculated from calvings before 1999 were compared to February 2004 breeding values (BV) calculated from all records or from only first parity records. For bulls born since 1990 with reliabilities above 60% for the current evaluation, within birth year correlations were 0.01 to 0.02 higher for predicting first parity BV from a model with 8% of variance assigned to maturity effects than from a model without this effect. For most birth years after 1989, correlations were not higher for predicting the all parity BV from models with maturity effects as compared to the current repeatability model. A range of maturity variances from 2 to 16% was tested. If 8% of phenotypic variance was assigned to maturity effects, genetic correlations of first with second through fifth parity were assumed to be 0.89, 0.87, 0.87, and 0.87. Correlations among third through fifth were 1 and second with later were >0.99. This random regression approach accounts for genetic differences in rate of maturity with little computational expense, but most genetic parameters tested did not result in improved evaluations.