We introduce a hierarchical spatio-temporal regression model to study the spatial and temporal association existing between health data and air pollution. The model is developed for handling measurements belonging to the exponential family of distributions and allows the spatial and temporal components to be modelled conditionally independently via random variables for the (canonical) transformation of the measurement mean function. A temporal autoregressive convolution with spatially correlated and temporally white innovations is used to model the pollution data. This modelling strategy allows first to predict pollution exposure for each district and then to link them with the health outcomes through a spatial dynamic regression model.
This talk refers to a joint work with Lara Fontanella, Clara Grazian and Luigi Ippoliti