Data consisting of 9,374 test day milk yield records from 302 first lactations of Alpina goats were analyzed by random regression models using the Wilmink and Ali and Schaeffer functions and Legendre orthogonal polynomials of third and fifth orders. Models including animal additive genetic, permanent environmental and homogeneous or heterogeneous (three or four classes) residual random effects were compared by Akaike information criterion (AIC), Schwarz Bayesian information criterion (BIC), likelihood ratio test (Ln L), phenotypic, permanent environmental, genetic and residual variances and by heritability estimates. According to AIC, BIC and Ln L, Legendre orthogonal polynomial of fifth order with three or four residual classes and Ali and Schaeffer function with four residual classes were the best fitting models. These models differed by the partition of phenotypic, permanent environmental, genetic and residual variance estimates in the beginning and in the end of the lactation period. Genetic correlation estimates between milk yields in the beginning and in the end of lactation obtained by Ali and Schaeffer function were negative. Legendre polynomial of fifth order assuming heterogeneous residual variance was the best fitting model for test day milk yield of Alpina goats.
Ali and Schaeffer function; Legendre orthogonal polynomial; random regression; Wilmink function
