Records of 11,023 first-parity Holstein cows belonging to 251 herds in the State of Minas Gerais were used to compare the Legendre polynomials and Wilmink function in random regression models (RRM) as for their effects in the estimate of genetic parameters and prediction of breeding values for nine types of persistency measurements and 305-day milk yield. The random regression test day models included the effect of herd-year-month test day, parameters of the function of Wilmink or 3th to 5th order Legendre polynomials to model fixed curves of the subclasses and 3th to 5th order Legendre polynomials to model genetic and permanent environmental effects. The Akaike’s Information Criterion (AIC) and Bayesian Information Criterion (BIC) indicated the model with larger number of parameters as the one that best fitted the data of milk yield. Using the Legrendre polynomial model large variation was observed in the estimates heritabilities for most of the persistency measures. The estimates herdabilities varied from 0.11 to 0.33 to milk yield throughout the lactation, from 0.33 to 0.36 for the 305-day milk yield and, from 0.00 to 0.32 for persistency. Genetic correlations between persistency and 305-day milk yield differed according to the model and persistency measure. Compared the Legendre polynomials to the Wilmink function provided expressive changes in rank of animals for persistency of lactation.
random regression model; genetic parameters; test day; Holstein cows; breeding value
