Study of milk yield and fat yield curves of some Iranian buffalo breeds

In order to study milk and fat yield curves of Khuzestani and Azeri breeds in 1st lactation, a total of 22.731 and 21.571 test-day records for each trait were used in the analyses. Eleven mathematical functions were used for fitting the curves. Adj, R2AICc, and Durbin Watson statistics were used for evaluation of the functions. A function with higher Adj-R2 but lower AICc and without auto correlated residuals is more desirable. According to Adj-R2, Diphasic and Rook functions had the highest (0.94) and the lowest (0.38) values in Khuzestani breed, respectively. Whereas, most of the functions had similar results in Azeri buffaloes, except for Legender polynomial (0.92), Monophasic (0.91), and Rook (0.88) functions that had lower values of the Adj-R2. Evaluation of the functions based on AICc statistic in Khuzestani breed showed that Diphasic and Legender polynomial functions were the best in fitting the milk yield and fat yield curves, respectively. However, Rook and Cobby functions acquired the lowest ranks in fitting the traits’ curves. Also, Mixed log and Ali & Schaeffer functions were among the ones with highest ranks in Azeri breed. Rook and Monophasic functions, however, had the lowest ranks. According to Durbin Watson statistic, residuals of Diphasic, Legender polynomial, Dijkstra, and Ali & Schaeffer functions were not positively auto correlated. However, residuals of the other functions were, in some cases, positively auto correlated. None of the functions in Azeri breed showed auto correlated residuals. In Khuzestani breed, however, most of the functions showed positively auto-correlated residuals.