New statistical models for genetic parameters estimation based on longitudinal data have been proposed. In this study, we considered the data of 47,614 test-day milk yields from 1,578 buffaloes, with 4,757 complete lactations, calving from 1985 to 2006. Single- ,two-and multiple trait analyses were used and variance components were estimated by Restricted Maximum Likelihood. The model used to evaluate the milk yield trait at 305 days (MY305) included: i) herd-year-calving season (contemporary group) and milking number as fixed effects; ii) buffalo age at calving as covariate (linear and quadratic effect); iii) additive genetic, permanent environmental, residual and the animal as random effects. The same effects were included in the test-day milk yield model except the contemporary group, defined as herd-year-test day. The heritability estimations for TDMY vary from 0.13 to 0.23 for single-trait analyses, from 0.13 to 0.24 for two-trait analyses and from 0.15 to 0.24 for the multiple-trait analyses. The results obtained for each of the three models showed that the higher heritability estimations were always obtained in the third test month. The genetic correlations between the TDMY and MY305 were high and positive. In conclusion, the TDMY trait use into the selective process will promote genetic changes similar to the ones obtained through the MY305.Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)