Predicting egg production in chukar partridges using nonlinear models and multivariate adaptive regression splines (MARS) algorithm [Vorhersage der eiproduktion bei chukarhühnern mit nichtlinearen modellen und multivariate adaptive regression splines (MARS)-algorithmen]

Abstract

The aim of this study was to compare Grossman-Koops, cubic and segmented polynomial models with an alternative non-parametric regression technique called multivariate adaptive regression splines (MARS) algorithm for predicting egg production in the Chukar partridge. The data on the number of eggs of 52 partridges over the laying period were collected twice a day for 23 weeks. Model fit statistics such as coefficient of determination (R2), adjusted coefficient of determination (Adj. R2), mean square error (MSE), root mean square error (RMSE) and Akaike’s information criterion (AIC) were used to measure the predictive abilities of the fitted models. The MARS was found as the best model defining egg production of the Chukar partridges with the highest R2 and Adj. R2 values (0.995 and 0.996) and the lowest MSE, RMSE, and AIC values (2.057, 1.434 and 41.000, respectively) followed by the segmented polynomial model. The results of the study denoted that the MARS predictive model with an easier formula and higher accuracy can serve as a better alternative to classical non-linear models in predicting cumulative egg production. © Verlag Eugen Ulmer, Stuttgart.