Comparison of different growth curve models in romanov lambs

Abstract

This study investigates the modelling of the individual growth curves of Romanov lambs using different equations and the data on the increase in live weight and selects the best model. For this purpose, the live weights of Romanov lambs that were brought to Gürçeşme Village, Niksar, Turkey, from Nikopol, Russia, were recorded from birth to day 180. In the study, the Cubic Spline, Logistic, Gompertz and Richard models were used. For the study, individual growth curves of a total of 278 (178 females, 100 males) lambs were modeled. For the selection of the best model, adjusted determination coefficient (R2 adj.), mean square error (MSE), Akaike information criteria (AIC) and Durbin-Watson (DW) values were used. In addition, attention was paid to the parameters and standard errors of the models. The results showed that the mean square error for the male lambs varied from 0.295 to 0.995, while it varied from 0.995 to 2.659 for the female lambs; the R2 adj. values were between 0.971 and 0.997 for the male lambs and 0.969 and 0.993 for the female lambs. The AIC values were between -37.12 and 0.094 for the male lambs and -0.196 and 122.12 for the female lambs. The DW values ranged from 1.86 to 2.44 for the female lambs and from 1.02 to 2.79 for the male lambs. Considering the MSE, R2 adj., AIC and DW values of the female lambs (0.295±1.195, 0.997±0.002, -37.12±0.001, 2.23±0.49, respectively) and male lambs (0.995±1.021, 0.993±0.001, -122.12±0.05, 2.31±0.19, respectively), the Cubic Spline model was determined to be the best model, while the Richard model was determined to be the worst fitting model both for the female (0.95±5.143, 0.971±0.002, 0.094±0.31, 2.41±0.01) and male (1.85±2.569, 0.969±0.011, -0.196±0.04, 2.79±0.05) lambs.

Keywords

• romanov
• lamb
• modelling
• growth curve models