Uma proposta para análise de dados com correlação espacial e temporal

Abstract

In several research areas, the study of the occurrence of a phenomenon over a period of time is very common. In this case, if we use the theory of Generalized Linear Models to analyze the subject of interest, we ll have, as a consequence, incorrect inferences concerning regression parameters and inefficient estimators, since considering the random variables as independent responses, is the main characteristic of this theory. When the variable response is observed over time, there can be a correlation between the observations and that must be taken into consideration on the estimation of the parameters. To incorporate this temporal dependence, we can use the theory of Generalized Estimating Equations, proposed by Liang & Zeger, 1986, as an extension of the Generalized Linear Models, to compute the correlation between the observations. Besides the temporal correlation, there can even be a space correlation and, in this case, we can use the theory of Geostatistic to estimate the reach of the correlation of samples over a study region, as well as to identify whether there is a privileged direction of variability of the studied phenomenon, important data not revealed when we utilize the theories of classic statistic. In this dissertation, we applied the methodologies mentioned above to try to explain the presence and the behavior of the female Aedes (Stegomyia) aegypti captured by adulticed traps in the city of Mirassol/SP, with the goal of helping on the search of more precise methods to contain the dissemination of dengue.