Comparison of a genetic group and unknown paternity models for growth traits in Nellore cattle.

The aim of the present study was to compare a model assuming unknown paternity and a model using genetic grouping to indicate the most adequate statistical procedure for the estimation of breeding values for animals with uncertain paternity. After data consistency, 62,212 Nellore animals, offspring of 581 bulls and 27,743 cows, were used in the analyses. The pedigree file contained 75,088 animals, including 22,810 (30.18%) offspring of multiple sires and 12,876 animals belonging to the base population with unknown parents. Three different approaches were adopted to deal with uncertain paternity of multiple-sire (MS) offspring. In the model of unknown paternity, the MS groups were ignored, and the sires of MS offspring were considered to be unknown and to belong to a single base population. In the genetic group approach, 2 definitions were used. In the first definition (GGa), "phantom parents" for animals with uncertain paternity were attributed, defining the genetic group as the MS group. In the other approach, GGb, phantom parents for animals with uncertain paternity were also attributed; however, MS offspring were clustered in genetic groups according to their year of birth, every 3 yr, on the basis of the average of male generation interval. Univariate analyses were performed under the Bayesian approach via Markov chain Monte Carlo methods. Models were compared by deviance information criteria and the conditional predictive ordinate. According to the choice criteria results, the genetic group model defined by the generation interval of males was more appropriate for predicting the genetic merit of animals with uncertain paternity. Therefore, the use of this model is recommended for the prediction of genetic merit and classification of offspring of multiple sires.