The association between the daily weight of the sediment
(smoke particles and other particles) and the daily
number of visits was evaluated by means of Pearson correlation
coefficients. We considered the daily number of
visits as the dependent variable in generalized additive
Poisson regression models. A robust technique (M-estimation)
was used to control for the influence of outliers.
Smooth functions (loess) were used to control for time
trend and temperature, and linear terms were used for
days of the week. When necessary, autoregressive terms
were included in the models to control for autocorrelation
of residuals. The best model was the one that presented
the lowest AIC (Akaike’s Information Criteria). To minimize
the influence of extreme points of the dependent
variable, we considered in the models only the counts of
inhalation therapy in the range between 5 and 95%.