Use of multivariate adaptive regression splines for prediction of body weight from body measurements in Marecha (Camelus dromedaries) camels in Pakistan

The aim of the present work was to predict live body weight by means of some body measurements, i.e., SH, CG, and BG in indigenous Marecha camel breed. For this purpose, multivariate adaptive regression splines (MARS) algorithm was used at proportions of various training and test sets, i.e., 65:35, 70:30, and 80:20 in V-tenfold cross-validation. In prediction of live body weight of the Marecha camels (160 female and 145 male animals) in the MARS predictive models, pairs of sex-SH (model 1), sex-CG (model 2), and sex-BG (model 3) as potential predictors. The best MARS model in LW prediction was obtained using sex and SH independent variables for 80:20 training and test set. Sex was determined to be an important source of variation in SH, CG, and BG as a result of sexual dimorphism in camels (P < 0.01). MARS results indicated that SH could be used as an indirect selection criterion to obtain elite camel herds on LW of Marecha camels. If genetically confirmed, the Marecha camels whose SH is taller than 165.1 cm could be selected for providing genetic progress in LW. In conclusion, use of MARS algorithm may be worthy of consideration for better identification of camel breed standards and selection of superior Marecha camels for meat productivity in Pakistan.