Proposição de novas metodologias para análise de aleatoriedade em processos pontuais no espaço-tempo

Abstract

The information on the spatiotemporal dynamics of plant diseases is of paramount importance to develop appropriate technologies to the management of diseases in production systems. Thus, in recent years, several studies involving spatial and temporal patterns are being conducted to get a better understanding of the mechanisms and dynamics of diseases in agriculture. However, the analysis of the space-time interaction, using point processes methods, has been scarcely discussed in this area. The aim of this paper is to present methods based on both the second-order statistic and the scan statistic to analyze point processes to detect the presence of clusters in time, space as well as clusters from the spatial and temporal interactions. The paper also proposes a new hypothesis test to detect clustering in time, using the distribution function of time counts of the occurrence of events, and a hypothesis test to detect spacetime interaction. The two hypotheses tests are based on Monte Carlo techniques. The methods presented were tested in the four realizations generated through computer simulation and applied to monthly data on incidence of Citrus Sudden Death (SCD) in orange. The results from analyzes of both simulated and actual data from SCD showed that the methodologies available in the literature and suggested in this thesis were effective to detect clusters in time, in space and space-time.