O modelo aditivo generalizado e a técnica de Bootstrap: uma associação entre o número de atendimentos hospitalares por causas respiratórias e a qualidade do ar

Abstract

The generalized additive models (GAMs) have become reference in the analysis of short-term effects of air pollution on human health. Lately, an effect called GAM concurvidade (analogous to multicollinearity in parametric modeling) has been found. This effect leads to underestimation of standard errors affecting its asymptotic confidence intervals of the parameters of the model. Some studies have proposed the use of the conditional bootstrap method to construct confidence intervals for the model parameters. This dissertation uses the GAM and the bootstrap techniques to explain the association between the number of hospital visits due to respiratory diases in children age from 0 to 6 years old and daily concentrations of pollutants (PM10, O3 and NO2). The results show that, in general, the procedures and conditional bootstrap confidence intervals have satisfying performances when used in GAM. The GAM confirms a relation between air pollution and children’s health conditions during the studied period.