Ajustes de curvas de lactação de vacas mestiças das raças holandesa e gir.

Abstract

The object of this study was to adjust the lactation curve of a cattle herd, using 415 lactations, including 19,025 controls of 206 crossbred cows of Holstein and Gir. Data from weekly milk weights used in this study are from property located in the municipality of Beberibe, in the littoral zone of Ceará State, in the period from 2006 to 2010. The adjustment was made for the average lactation curve, using six linear mathematical functions: Linear (FL), Quadratic (CF), Quadratic Logarithmic (FQL), logarithmic (FLG), hyperbolic (FH) and Cubic (FC), two intrinsically linear: Incomplete Gamma (FGI) and Exponential (FE) and five nonlinear described by Papajcsik & Bodero, 1988 (FPD); Ferrel & Jenkins, 1984 (FJF); Dhanoa et al., 1981 (FD); Borlino-Cappio et al., 1995 (FCB) and Sikka, 1950 (FS). The criteria used to assess the quality of fit for each function were the adjusted coefficient of determination (R²a), the mean square (QMR), the Durbin-Watson and distribution chart of residue and comparison of curves and estimated observed by visual evaluation. The values of R²a for linear models ranged from 27 to 99%, with 81% of the R²a the estimates presented high values for the functions set, indicating that these features could be used in adjusting the lactation curve mean that flock. While models for intrinsically non-linear models FGI the estimated values of R²a greater than 80% except for the genetic group 1/4H showing values of 39%. The QMR ranged from 0.60 to 1.518 and showed a better fit for all GG. Since the nonlinear function that best fits was Dhanoa et al., 1981 (FD) had values of R²a to high ranging from 39 to 99% for the genetic groups. However, there was no tendency of correlation between successive residues for these functions, since 83% of DW values were higher than 2.