Matrizes socioeconômicas no ajuste de modelos STARMA aplicados a dados epidemiológicos

Abstract

In this work the use of socioeconomic neighborhood matrices was studied in time-space models of autorregressive and moving averages (STARMA) class. The selected data set is composed of nine time series that quantify the incidence rate of Tuberculosis observed between 2002 and 2017 in the following cities: Belo Horizonte, Betim, Contagem, Governador Valadares, Juiz de Fora, Lavras, Montes Claros, Pouso Alegre and Uberlândia. Since most cities are geographically distant, the use of socioeconomic neighborhood matrices was necessary. The matrices were obtained through two socioeconomic variables: the municipal IDH and the average annual investment in basic health. The model was obtained computationally and consisted of three stages: Identification, estimation and diagnosis of the model. It was concluded that, contrary to the imagined, it is possible to observe the existence of space-time autocorrelation in the incidence rate of tuberculosis, even in cities that are geographically distant. The distance between the areas observed in this work has made the socio-economic neighborhood matrices become the most appropriate option in the adjustment of STARMA models to the data used in this work.