Predictive ability of boiler production models

Abstract

The weekly body weight measurements of a growing strain of Ross broiler were used to compare the of ability of three mathematical models (the multi, linear, quadratic and Exponential) to predict 8 week body weight from early body measurements at weeks I, II, III, IV, V, VI and VII. The results suggest that the three models fitted the 8th week body weight records at the seventh-week body measurement closely, judging from their respective R2-values (linear, 0.58; quadratic, 0.58 and exponential, 0.57). However, their ability to predict 8th-week body weight from early records using R2-value, varied generally from 0.15 + 0.36 to 0.58 + 0.23 for all the models. The predictions of 8th-week body weights based on fit to 7th-week data were more accurate for the linear and quadratic models, though all the models predicted less than the actual 8th-week records (-32, -35 and -38.14%) for linear, quadratic and exponential respectively. It could be concluded that based on the goodness of fit to 7th week body measurements record accuracy to predict 8th-week body weight from early record, the linear model gave the best result followed by the quadratic and exponential respectively. It could therefore be said that the models have graded theoretical advantage. This may be found useful in decision making concerning replacement of broiler chicken.

Keywords: Poultry, broiler production, modeling, predictive ability