Abstract
A total of 5170 weekly average milk records were collected during the period from May 2011 till May 2013, from 127 Ardi
does, a native breed of the Kingdom of Saudi Arabia. The objectives were to describe the shape of lactation curve of milk
production by different functions, and to compare the results obtained from fitting these models. Four mathematical models were fitted and compared, namely the linear (LW) and nonlinear (NLW) forms of the Wood, the monophasic (MONO) and the diphasic (DI) functions. Comparisons of these models were carried out based on root mean square error (RMSE), Akaike Information Criteria (AIC), Bayesian Information Criteria (BIC) and Durbin-Watson statistic (DW). The results
indicated that the DI fitted the data better than the other models, as judged by lower mean square prediction error. The
predicted milk yield estimated by DI model was markedly closer in magnitude to the actual milk yield than those predicted
by the other models. Correlation coefficients between actual milk yield and predicted milk yield calculated by DI function
were higher than the corresponding coefficients estimated by the other functions. The estimates of DW statistic indicated
that the residuals obtained from fitting the LW, NLW and MONO functions were positively and highly correlated. The
corresponding values obtained by DI function were better. The AIC and BIC values obtained by fitting DI function were
larger than those estimated from fitting the other functions. In conclusion, DI function would be the model of choice for
describing lactation curve of Ardi goats. The ranking of the models should be taken with caution because of the small size
of the data set used in this study.
