Estimating the extent of degradation of ruminant feeds from a description of their gas production profiles observedin vitro: comparison of models

• DOI: 10.1017/s0007114500000179
• Publication Date: 2011-03-11T10:04:45Z
• Authors: M. S. Dhanoa; S. Lopez; J. Dijkstra; D. R. Davies; R. Sanderson; B. A. Williams; Z. Sileshi; J. France
• Keywords: Mathematical models; Likelihood Functions; Rumen; 0402 animal and dairy science; Ruminants; 04 agricultural and veterinary sciences; Animal Feed; Gas production; Feed degradation; Fermentation; Animals; Gases; Mathematical Computing

An evaluation of general models that describe gas production profiles is presented. The models are derived from first principles by considering a simple three-pool scheme and permit the extent of ruminal degradation to be calculated, as described in the companion paper. The models evaluated were the generalized Mitscherlich, simple Mitscherlich, generalized Michaelis–Menten, simple Michaelis–Menten, Gompertz, and logistic. Five sets of gas production data consisting of 216 curves, obtained using a wide range of feeds (including straw, hay, silage, grain and various byproducts), were analysed to study the performance of these gas production models. Application of the non-sigmoidal models (simple Mitscherlich and Michaelis–Menten) to the data resulted in convergence problems and these models were found to be inadequate in many cases. Based on results of a pairwise comparison between models (variance ratio test), ranking of residual mean squares, lack-of-fit test, and of analyses of residuals, the generalized Mitscherlich and the generalized Michaelis–Menten models seemed particularly suited because of their flexibility to encompass sigmoidal and non-sigmoidal shapes of gas production profiles, whether symmetrical or not.