The objective of this work was to estimate covariance functions for test-day milk yield (TDMY). Twenty-seven thousand TDMY records of 3,362 first lactations of Gir cows calving between 1990 and 2007 were analyzed. The TDMY data were grouped in 20 biweekly classes, analyzed by random regression models, whose random, additive-genetic and permanent environment effects were fitted by Wilmink's (W) or Ali & Schaeffer's (AS) functions. The residual variance (VR) modelling was fitted by 1, 4, 6, or 10 classes. The heritability estimates for TDMY varied from 0.12 to 0.32 for the AS function, and from 0.09 to 0.33 for W function, and were larger in the begining of the lactation period. The genetic correlations decreased from near unity, among adjacent TDMY values, to negative values between the first and the two last quarters of lactation. The model using the AS function with four VR classes, is a parsimonious option for the TDMY fitting of Gir cows in Brazil.
dairy cattle; parametric function; Bayesian inference; genetic parameters
