Abstract
This thesis aims to analyze a dataset containing 44 variables divided into 5 macro categories: identification variables, Air Quality (AQ) indicators, Weather (WE) variables, Emissions (EM) and Livestock (LI), and Land Use (LA). The objective is to estimate a GAMM model to describe the level of atmospheric particulate matter, taking into account the autocorrelation present in the residuals. This model will then be used for forecasting. The dataset was created as part of the AGRIMONIA environmental investigation project, which aims to deepen the understanding of the impact of agriculture on air quality in Italy.
In the first part, an exploratory analysis of the quantitative and qualitative variables was conducted. After analyzing them, it was found that most of them had non-linear relationships, and the GAMM model was used on the R software, which allows setting non-functional forms. Once the optimal model to explain particulate matter, considering the complex relationships between variables and between variables and particulate matter, had been identified, it was used to attempt to make forecasts.
|