Aplicação de modelos beta inflacionados de zeros para análise de dados longitudinais

Abstract

Zeros excess data are often found in practice, adequate analysis considers models that appropriately support, also, null observations counting. In this research, the Normal Random Effects Zero Inflated Beta Regression Model was studied and applied to modeling a experimental data set to describe the behavior, over time, of bacterial leaf citrus canker expected incidence in orange orchards, under influence of original rootstock and genotype. The adjustment allowed to quantify chances that a null mean incidence observation will come from a certain plant, according to rootstock and genotype, and to estimate its expected value according to this combination. The agronomic experiment considered all combinations between distinct four rootstocks and nine genotypes. The longitudinal model found that both, proportion of uninfected plants and expected incidence among diseased plants, underwent time action statistically significant influence, however the rootstock and genotype effect influence significantly just the proportion of uninfected plants. Laranja Caipira rootstock proved to be the most resistant to leaf citrus canker, as was the most susceptible is Limão Cravo rootstock. The genotypes IpiguaIAC, Arapongas, EEL and N59 presented the biggest, and statistically equivalent, chances of presenting an uninfected plant.