Milk production data of 19,151 lactations from 8,984 cows, daughters of 795 Gyr sires were used to genetically evaluate cows. Mixed model equation was used to predict sire-breeding values that were incorporated in the selection index of the cows. Besides the selection index, animal models were also used to predict the genetic value of the cows, using only the first lactation or up to five lactations per cow. The accuracies of the predictions obtained by cows breeding values, using the animal models, were higher when were used in the evaluation up to five lactations (.76) or only the first lactation (.73) on the accuracies predictions obtained by the selection index (.55 to .59). Similarly, the correlation among the breeding values predicted by selection index and predicted by animal model up to five lactations ranged from .74 to .84. The highest correlation (.84) was observed among the prediction obtained by animal models and the lowest (.74) within the animal model with up to five lactations and the selection index, which the sire-dam breeding values were obtained by the herdmate method. The genetic gain obtained when 20% of best cows were selected to produce descendants for the new generation was at least 80% of the gain obtained by animal model (270 kg), regardless the used procedure.
cow evaluation; correlation; methodology; accuracy
