Published online by Cambridge University Press: 18 August 2016
The objective of this study was to estimate genetic parameters and breeding values of sires for different definitions of length of productive life in Swedish Red and White dairy cattle. The data consisted of 534 016 daughters with initial calvings between 1988 and 1996. These cows were daughters of 1266 sires (55 of which were considered proven bulls, and treated as fixed effects). The model for the hazard included: a random time-dependent effect of herd-year-season (hys), a fixed time-dependent effect of year-season, a fixed time-dependent effect of parity by stage of lactation, a fixed time-dependent effect of the cow’s peak yield as deviation from her herdmates in that herd-year, a fixed time-independent effect of age at first calving, and random and fixed effects of sire. The hys effect was assumed to follow a gamma distribution and the random sire effect a normal distribution with mean zero and variance Aσ2s, where σ2s is the variance among sires and A the relationship matrix. Length of productive life was defined as the number of days from first calving to culling or end of data. Two types of length of productive life were studied: (1) functional productive life (PL), all cows that were culled before the end of data were considered as uncensored; (2) fertility determined productive life (FPL), only cows that were culled for fertility problems were considered as uncensored. Heritability estimates were 6·9% and 6·1% for PL and FPL, respectively. The approximated genetic correlation estimates between fertility measures in first lactation and FPL was higher than the corresponding estimates with PL. In conclusion, it was feasible to use the trait fertility-determined length of productive life, its heritability was almost as large as for PL, and the genetic correlation with fertility was higher than for PL. However, the expected selection response in fertility from indirect selection on FPL was lower than from direct selection.