Prediction of apparent metabolisable energy content of cereal grains and by-products for poultry from its chemical composition

In order to predict the metabolisable energy content of ninety batches of cereal grains and cereal by-products for poultry, regression models derived from different sample aggregations and using chemical components as independent variables were compared. Several statistics have been calculated to estimate the error of prediction. The results indicate that the highest levels of significance and coefficients of determination were obtained for equations derived from the larger data sets. However, the lowest prediction errors were associated to equations calculated for data or groups of data closer to the ingredient ­studied