Multiple Trait Random Regression Test Day Model for Production Traits

Research on test day (TD) models began in Canada in 1991. At that time fixed regressions (within age and season of calving) were used to account for the shape of the lactation curve of dairy cows, but the animal effect was used only to account for differences in height of these curves (Ptak and Schaeffer, 1993). Fixed regression TD model was applied to Canadian dairy goat data (Schaeffer and Sullivan, 1994) and to Canadian data on somatic cell score (SCS) (Reents et al., 1995a,b). Schaeffer and Dekkers (1994) suggested the possibility of using random regressions in a linear model (Henderson, 1982) for analysing TD data. Single trait random regression models were applied to first lactation milk, fat and protein records, with different functions for fixed and random regressions (Jamrozik and Schaeffer, 1997; Jamrozik et al., 1997a,b). In the simulation study of Kistemaker (1997) random regression models were significantly better than an analysis of 305d yields in terms of correlation between estimated and true breeding values. A 2-3% increase in accuracy for bulls and 68% for cows was found for first lactation milk yield. Changes occuring in Canadian milk recording system will require an application of multiple trait test day model in genetic evaluation for dairy production traits. Future milk recording programs would have TD records that have all yield values (milk, fat, protein, SCS) while other TD records might have only milk yield. Thus the multiple trait model for simultaneous analysis of all yields seems to be a logical choice. Also, preliminary work with milk yield in different lactations showed that lactation curves were different between first and second lactation, and again between second and later lactations. Thus, yields in different lactations should also be considered as different traits. The objective of this research was to develop the multiple trait, random regression TD model for genetic evaluation of dairy bulls and cows for production traits. Both animal genetic and permanent environmental effects were modelled by random regressions in this model. Some preliminary results are presented.