Using an algorithm to fit a gamlss model on dry matter data from brachiaria brizantha

Background and Objective: Forage production in the tropics is generally asymmetrically distributed. Hence the need to use more complex models, especially when multiple comparisons are made and there are very large deviations from normality. The objective of this research is to fit a Generalized Additive Model for Location, Scale and Shape (GAMLSS) model on accumulated dry matter data from Brachiaria brizantha using a model selection algorithm. Materials and Methods: A Box-Cox Power Exponential (BCPE) distribution was adjusted on the dry matter from Brachiaria brizantha data implementing GAMLSS in R (programming language). The accumulated dry matter data for B. brizantha were obtained from a study carried out on a farm in the state of Portuguesa, Venezuela. The explanatory covariate x was the interval between cuts (21, 28, 35 and 42 days). Results: The dependent variable (dry matter) exhibited both skewness and kurtosis. GAMLSS allowed flexible modeling of both the distribution of the dry matter yield from B. brizantha and the dependence of all the parameters of the distribution on intervals between cuttings. For the dry matter yield from B. brizantha, which exhibited skewness and leptokurtosis, the BCPE distribution, provided the best fit. Conclusion: The interval between cuttings showed an effect that is reflected in the average yield of dry matter from B. brizantha. The interval between cuts affected the skewness and the kurtosis of the distribution.