CONTROL EVALUATION OF CONYZA SPP. WITH GAMLSS MODELS: AN EFFICIENT ALTERNATIVE TO THE TRADITIONAL ANALYSIS OF VARIANCE MODEL

Resumen

The study aimed to present a framework on the fit of the Generalized Additive Models for Location, Scale, and Shape (GAMLSS) with the support of R software to evaluate the control of Conyza spp. in pre-sowing desiccation of soybean during an advanced phenological stage through herbicide associations and/or sequential applications. GAMLSS is an alternative to the traditional model, which allows modeling the response variable using a large class of probability distributions. The dataset used in this study refers to the control of Conyza spp., ranging from 0 to 100%. Visual evaluations were performed at 7, 14, 21, 28, and 35 days after application of different herbicides. The results of this study indicate that GAMLSS may be an efficient alternative to the traditional analysis of the variance model, especially when the data cannot be represented by a normal distribution or exhibit heteroscedasticity.