Author
SOUTHEY, B - UNIV. IL, URBANA | |
RODRIGUEZ-ZAS, S - UNIV. IL, URBANA | |
Leymaster, Kreg |
Submitted to: Journal of Animal Science Supplement
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 11/15/2001 Publication Date: 7/4/2002 Citation: Southey, B.R., Rodriguez-Zas, S.L., Leymaster, K.A. 2002. Estimation of genetic parameters of lamb mortality using discrete survival analysis. Journal of Animal Science Supplement. 80(Suppl. 2):43. Interpretive Summary: Lamb mortality from a composite population at the U.S. Meat Animal Research Center was studied using descrete time survival analysis since the actual time of mortality may be unavailable but information about a particular interval is available. Mortality records from 8,642 lambs were separated according to age into three stages, birth to weaning (BW), weaning to 365 d of age (WY) and birth to 365 d of age (BY). Within each stage animal-period records were constructed with period defined as weekly or fortnightly. In addition, a daily animal-period was studied for BW and a monthly animal-period was studied for WY and BY. The logistic and complementary log-log link functions were used in sire, animal, and animal and maternal effects models and compared to continuous time survival Weibull model. The estimates of the sire variance were very similar between link functions and length of animal-period in all stages and to estimates obtained using the Weibull sire model. The complementary log-log link function heritabilities (standard errors) were 0.18 (0.06), 0.20 (0.20) and 0.13 (0.04) for BW, WY and BY stages, respectively, and were very similar to those obtained using a Weibull sire model. However, these estimates were approximately twice the magnitude of the estimates obtained with a logistic link function. These differences are probably due to the different variances of the link functions. Estimates of additive genetic variance and the associated heritability obtained with an animal or animal and maternal effects models were lower than the sire model for the same scenario. These results illustrate that the use of discrete time approaches constitutes a feasible alternative to continuous time approaches in survival analysis. Technical Abstract: Lamb mortality from a composite population at the U.S. Meat Animal Research Center was studied using descrete time survival analysis since the actual time of mortality may be unavailable but information about a particular interval is available. Mortality records from 8,642 lambs were separated according to age into three stages, birth to weaning (BW), weaning to 365 d of age (WY) and birth to 365 d of age (BY). Within each stage animal-period records were constructed with period defined as weekly or fortnightly. In addition, a daily animal-period was studied for BW and a monthly animal-period was studied for WY and BY. The logistic and complementary log-log link functions were used in sire, animal, and animal and maternal effects models and compared to continuous time survival Weibull model. The estimates of the sire variance were very similar between link functions and length of animal-period in all stages and to estimates obtained using the Weibull sire model. The complementary log-log link function heritabilities (standard errors) were 0.18 (0.06), 0.20 (0.20) and 0.13 (0.04) for BW, WY and BY stages, respectively, and were very similar to those obtained using a Weibull sire model. However, these estimates were approximately twice the magnitude of the estimates obtained with a logistic link function. These differences are probably due to the different variances of the link functions. Estimates of additive genetic variance and the associated heritability obtained with an animal or animal and maternal effects models were lower than the sire model for the same scenario. These results illustrate that the use of discrete time approaches constitutes a feasible alternative to continuous time approaches in survival analysis. |