Accuracy of predicting genetic merit of A.I. sampled bulls from pedigree information and the impact of son's proof on dam's PTA

TR Number
Date
1992
Journal Title
Journal ISSN
Volume Title
Publisher
Virginia Tech
Abstract

A total of 1,644 A.I. sampled bulls born from 1984 to 1986 with first proofs from Winter 90 to Summer 91 were used to determine the accuracy of predicting DYD and PTA from different sources of pedigree information obtained before the bull had daughter information. Traits evaluated were milk, fat and protein. Pedigree sources considered were PA, PI, PTASIRE and PTADAM. Approximate weighted regression was used to determine which pedigree source predicted DYD or PTA with the highest accuracy (highest R²). For all traits, PA had a higher R² for DYD and PTA than PI. Regression coefficients were less than one for PA and PI. R² values for PA to predict first DYD milk, fat and protein were .17, .20 and .18, respectively. R² for PA to predict first PTA milk, fat and protein were .47, .54 and .49, respectively. Adding PTADAM to the model with PTASIRE resulted in a higher R² than the model with PTASIRE alone. As expected R² values were similar for PA and the model with PTASIRE and PTADAM. However, the weights for PTASIRE and PTADAM were less than .5. Higher weights and R²s for predicting PTA compared to predicting DYD resulted from the part-whole relationship between bull’s PTA and his PA. Overall, weights and R² were less than expected, but reasonable accuracy was obtained in estimating a young bull’s DYD and PTA from pedigree estimates. Accuracy of prediction varied depending on when the bull received his first proof. R² values of different groups of bulls based on the date of first DYD and PTA ranged from .06 to .20, .08 to .15 and .05 to .12 for predicting first DYD from PA for milk, fat and protein, respectively. Prediction accuracy in some groups of bulls was less possibly because of the limited number of sires and reduced variation in sire PTAs. Changes in evaluation procedures to expand the variance of extended records and to account for differences in within herd variance may have adversely affected the accuracy of prediction.

The impact of the addition of granddaughters (son’s daughters) on the PTA of the dam was evaluated. Addition of granddaughter information decreased the average of dam’s PTA 70 kg, indicating the dams’ PTAs were generally inflated. Granddaughter information measured relative to PA of the son was useful to predict the change in the dam’s PTA at the AM evaluation the dam’s sons received first proofs. Regression coefficients ranged from .30 to .39, which were similar to the weights for w₃ in the PTA function. R for the regressions ranged from .33 to .72. Predicting further change in dam’s PTA (after the AM evaluation first granddaughter information was received) resulted in lower R? (.13 to .35) for additional granddaughter information.

Evidence of bias and/or errors were found in bulls sampled outside the respective A.I. organizations’ designated sampling herds. These bulls had PAs that overestimated their DYDs for milk, fat and protein by 107 kg, 7.5 kg and 5.7 kg, respectively. The PAs of these bulls overestimated the PTAs by 97 kg, 6.8 kg and 4.5 kg for milk, fat and protein, respectively. Discrepancies were also found between average PTAs and DYDs and the PAs of bulls based on the rank of the dam’s PTA. Bulls from dams with lower PTAs tended to have PAs that underestimated their DYDs by 48 kg and .5 kg for milk and fat, respectively. These bulls had PAs that underestimated their PTAs for milk, fat and protein by 42 kg, .5 kg and .6 kg, respectively. Examination of bulls from high ranking dams for PTA milk, fat or protein revealed that bulls from dams with higher PTAs tended to have PAs that overestimated their DYDs by 65 kg, 5.3 kg and 4.5 kg for milk, fat and protein, respectively. The PAs of these bulls overestimated their PTAs by 49 kg, 4.2 kg and 2.9 kg, for milk, fat and protein, respectively.

Description
Keywords
Citation
Collections