Modelling of dry matter yield of Pennisetum purpureum cv. king grass with different cutting frequencies and N fertilizer dosages

Abstract

For predicting the productive performance of Pennisetum purpureum cv. king grass with different cutting frequencies and N fertilizer levels, a non-linear mixed regression model and another of categorical regression were adjusted. Data from a five-year experiment were taken in which the effect of cutting frequency and nitrogen dosage (N) on DM yield was evaluated. Literature data from the last 30 years were collected for selecting the variables of highest influence on DM yield. The Bootstrap re-sampling was used for generating a sample of 50 observations per cut. Five statistical criteria were assessed for selecting that of better adjustment in the variable studied. It was established that the non-linear mixed regression model was not adequate for predicting the performance of DM yield, since it showed coefficients of determination lower than 80 %. The categorical regression achieved to predict DM yield with high precision in the adjustment, according to the established statistical criteria that allowed identifying the variables and effects of greater importance in the rainy and dry periods. The cutting frequency was the one of highest incidence in both periods.
Key words: non-linear, mixed effects, Bootstrap, DM yield