The objective of this research was to present a methodology for identification and modeling of residual autocorrelation considering individual adjustments of the Wood's model to lactation dairy goats and evaluate the influence of such modeling in the quality of adjustment. The Wood's model was adjusted individually for lactations in three different ways, the first have assumed independence of errors (IE) for all lactations, the second have assumed autoregressives first order errors (AR1) for all lactations and the third, named (IE-AR1), was used the AR1 errors structure only for lactations that showed residual autocorrelation according to Durbin-Watson test, and the IE errors structure for the other lactations. The three ways of adjustment were compared by the percentage of convergence and the average of the mean square errors (MSE) and coefficients of determination adjusted (R²adj). The average of MSE and R²aj were very similar in the three cases of residual structure. However, the model with IE-AR1 residual structure showed a higher rate of convergence, which is an advantage, as it allows a greater number of animals are evaluated for their lactation curve. Therefore, due to the increasing convergence obtained, the fit of the Wood's model with IE-AR1 residual structure is the option most suitable for large data sets.
autoregressives errors; Durbin-Watson; independent errors
