72
Evaluation of Different Gas Production Models When Applied to Ruminants Feeds
Evaluation of Different Gas Production Models When Applied to Ruminants Feeds
Monday, March 12, 2018
Grand Ballroom Foyer (CenturyLink Convention Center)
The objective of this study was to evaluate the goodness of fit of several nonlinear models to gas production profiles across different categories of ruminant feeds. In vitro incubations were completed for 48 h using rumen fluid from a lactating cow fed a 50:50 forage to concentrate diet. Gas production was recorded continuously every 5 min using an automatic system.
Gas production profiles of 101 feeds grouped into different categories; corn silage (n=15), small grains silages (n=15), alfalfa hay (n=11), alfalfa haylage (n=4), byproducts (n=20); protein meals (n=7), energy feeds (n= 17), and lactation TMR (n=12) were fitted to 8 models; exponential, Logistic, dual pool logistic, France, Gompertz, dual pool Gompertz, DeGroot, and McDonalds-Orskov. Models were evaluated based on three statistics including mean square error (MSE), coefficient of determination (R²), residual mean absolute deviation (RMAD) for each gas production fitting, and relative efficiency. These data were analyzed using Proc Glimmix of SAS (SAS Institute, 2009) and means were compared using Tukey test.
France and logistic models resulted in the highest MSE, and lowest R2 while other models had similar values (67.5 ± 9.84 and 0.969±0.015 vs. 11.0±4.51 and 0.993 ±0.0026), respectively. The dual pool Gompertz model had a relative efficiency greater than 1.0 when compared to all other models followed by Mc-Donalds-Orkov. When models were evaluated within each feed category, Gompertz was the best fit for byproducts, exponential for energy feeds, protein and small grain silages, dual pool Gompertz for grass hays and TMR, and dual pool logistic for corn silage and alfalfa hay. Although dual pool models may better fit gas production data for a wide range of feeds, specific models may be required for the best fit of a specific category of feeds.
Gas production profiles of 101 feeds grouped into different categories; corn silage (n=15), small grains silages (n=15), alfalfa hay (n=11), alfalfa haylage (n=4), byproducts (n=20); protein meals (n=7), energy feeds (n= 17), and lactation TMR (n=12) were fitted to 8 models; exponential, Logistic, dual pool logistic, France, Gompertz, dual pool Gompertz, DeGroot, and McDonalds-Orskov. Models were evaluated based on three statistics including mean square error (MSE), coefficient of determination (R²), residual mean absolute deviation (RMAD) for each gas production fitting, and relative efficiency. These data were analyzed using Proc Glimmix of SAS (SAS Institute, 2009) and means were compared using Tukey test.
France and logistic models resulted in the highest MSE, and lowest R2 while other models had similar values (67.5 ± 9.84 and 0.969±0.015 vs. 11.0±4.51 and 0.993 ±0.0026), respectively. The dual pool Gompertz model had a relative efficiency greater than 1.0 when compared to all other models followed by Mc-Donalds-Orkov. When models were evaluated within each feed category, Gompertz was the best fit for byproducts, exponential for energy feeds, protein and small grain silages, dual pool Gompertz for grass hays and TMR, and dual pool logistic for corn silage and alfalfa hay. Although dual pool models may better fit gas production data for a wide range of feeds, specific models may be required for the best fit of a specific category of feeds.