J.B. Cole1, D.E. Franke1, and E.A. Leighton2
1Department of Animal Science, Louisiana State University, Baton Rouge, LA 70803
2The Seeing Eye, Inc., Morristown, NJ
Data on longevity for 1,304 German Shepherd dogs (GS) was used to estimate genetic parameters for working life. A Cox proportional hazards model on unadjusted working life was used to test the assumption that the baseline hazard function for the population is a Weibull hazard. The rejection of that assumption led to the definition of two measures of working life: estimated for working life to 18 months post-graduation (EWL) and working life beyond 18 months post-graduation (LWL). For EWL, 92.45% of the records were censored after 540 days, while for LWL 47.89% of the records were censored after 4,361 days. The Survival Kit v3.12 was used for variance components and breeding value estimation. Estimates of additive sire genetic value for Cox and unstratified Weibull models were 0.87 and 0.36 for EWL and 0.05 and 0.03 for LWL, respectively. Heritability estimates on a log scale (assuming no censoring) were 0.19 and 0.61 for EWL and 0.09 and 0.04 for LWL, respectively. For both traits, the standard deviations and skewness of the posterior densities of the sire variance were fairly large. Due to the small size of the dataset and the large number of censored records, it was not possible to obtain more precise estimates of the sire variance. The large heritabilities observed for EWL under the Cox and Weibull models should be interpreted with caution because virtually all EWL records were censored. Pearson’s product-moment correlation coefficents for sire breeding values estimated under the Cox and Weibull models were 0.99 for EWL and 0.88 for LWL, respectively. The low correlation between the models for LWL may be due to large prediction error variances and the skewness of the posterion distribution of sire effects, although these problems were noted for EWL as well.
(Key words: variance components, working life, dog guides)