Fit of different functions to the individual deviations in random regression test day models for milk yield in dairy cattle

The shape of individual deviations of milk yield for dairy cattle from the fixed part of a random
regression test day model (RRTDM) was investigated. Data were 53,217 TD records for milk yield of 6,229 first lactation
Canadian Holsteins in Ontario. Data were fitted with a model that included the fixed effects of herd-testdate,
DIM interval nested within age and season of calving. Residuals of the model were then fitted with the following
functions: Ali and Schaeffer 5 parameter model, fourth-order Legendre Polynomials, and cubic spline with
three, four or five knots. Result confirm the great variability of shape that can be found when individual lactation
are modeled. Cubic splines gave better fitting pe4rformances although together with a marked tendency to yield
aberrant estimates at the edge of the lactation trajectory.