Modelos multivariados de análise discriminante como ferramenta de previsão de epidemias de dengue

• Autor(a) principal: Nóbrega, Diogo Medeiros
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UEPB
• Texto Completo: http://tede.bc.uepb.edu.br/jspui/handle/tede/3476
• topic: Software SPSS; Saneamento básico; Análise discriminante; Dengue; Discriminant Analysis; Basic Sanitation; Dengue; CIENCIAS DA SAUDE::MEDICINA

This research deals with the use of Discriminant Analysis and all its assumptions from variables related to basic sanitation actions, linking them to the incidence of dengue cases in the Brazilian States from 2013 to 2017 to model a discriminant prediction function capable of classifying the states into low-, medium- and high-level groups of dengue incidence. For the analyses, SPSS software was used. Epidemics generate high costs for health services, and the precarious infrastructure of public sanitation may directly contribute to growth in dengue cases. The use of statistical methods makes the investigation processes more direct and accurate, which makes it possible to perceive worrying circumstances and, thus, to anticipate and act for its control. In this work, the theoretical statistical development and the analyses are described. The features were extracted from epidemiological bulletin of the Ministry of Health and the time series of basic sanitation actions of the Ministry of Cities of the last five years available. It was identified that the variables Sewage Collection Index (ICE), Quantity of Municipalities Served with Sanitary Sewage (QMAES), Total Expenditure with Sanitation including water and sewage (DTS) and Quantity of Municipalities Served with Water Supply (QMAAA) were the most statistically significant. Finally, it was found that the separation between the low, medium and high incidence groups was statistically significant according to the tests, and the discriminant function reached a level of accuracy of 86.9% of the original High Incidence cases in their respective groups.