Analise da curva de crescimento de machos Hereford considerando heterogeneidade de variâncias e autocorrelação dos erros

Abstract

A weight-age growth curve was studied in 160 Hereford males. It was adjusted by using the Brody, Gompertz, Logistic, Richards and von Bertalanffy functions. The parameter estimates for the non-linear regression model were calculated by the ordinary minimum square and generalized minimum square methods (MODEL, SAS/ETS, 1995). The animal’s individual and mean adjustments were obtained by using the following models: weighing the inverse of variance with and without autoregressive error (AR) and by non-weighing with and without AR. The comparison among the above models was made by the parameters biological interpretations and by the quality adjustment evaluators (Adjusted determination coefficient, error mean square, mean predicted error, student's t-test for the parameters A and K estimates and convergency percent in the case of individual adjustments). The best fitted model was of that weighted by the inverse of variance with AR. The functions that best fitted the data were those of Gompertz and von Bertalanffy followed by Logistic. Brody and Richards functions presented a low convergency percent, being shown inadequate to describe the data in study. The individual fittings were more precise than the mediun fittings.