Determination of the most appropriate regression method to estimate metabolic energy values by in vitro gas production technique

Abstract

Estimation of metabolic energy values of feeds using in vitro gas production values is an essentialissue in practice. It is determined that estimation of metabolic energy values with lonely use of in vitro gasproduction was not achieved in previous studies. In this study, to determine the most fitted regression model forestimating metabolic energy with use of gas production in certain incubation periods and parameters of gasproduction equation, models were set taking into account distributions of data and models were compared withthe criteria of determination coefficient and mean square error. It is indicated that in vitro gas production valueshave long tailed, skewed, distributions such as Burr, Dagum and Generalized Pareto distribution. As a result, fivedifferent models with high coefficient of determination were set to estimate metabolic energy from in vitro gasproduction using normalized data. The model which estimates the metabolic energy with use of only gasproduction at 24 hours was determined as the best model from these five. Models were simplified to provideconvenience to users when they use models to estimate metabolic energy.