Summary: | Dissertation presented as the partial requirement for obtaining a Master's degree in Statistics and Information Management, specialization in Information Analysis and ManagementA criminalidade é um dos fenómenos sociais mais problemáticos que tem vindo a ganhar relevo e especial importância desde o final do século XX. É um fenómeno dinâmico e distribuído segundo tendências espaciais. Este tema tem sido objeto de estudo em diversas áreas científicas, ao longo do tempo, com o intuito de identificar novas leituras de informação, que possibilitem compreender os fatores associados à criminalidade, e encontrar soluções para prevenir e combater este problema. O evento criminal, tal como outros, desenvolve-se em torno do território, tendo por isso um cariz geográfico. Assim, a gestão e análise da criminalidade com recurso a ferramentas de informação geográfica têm representado uma importante contribuição para a análise da criminalidade. Esta investigação explora a distribuição espacial dos crimes contra o património em Portugal Continental e tem como objetivo principal investigar fatores demográficos e socioeconómicos, que possam estar associados a este fenómeno em cada município. Para tal, são aplicadas ferramentas de Sistema de Informação Geográfica (SIG), técnicas de análise exploratória, estatística descritiva e modelação espacial. A análise exploratória inicial permitiu identificar a existência de autocorrelação e heterogeneidade espacial na taxa de crimes contra o património, justificando assim a utilização de modelos de regressão espacial. Foram aplicados e comparados entre si, modelos de regressão de Poisson, nomeadamente o modelo global, o modelo local GWPR – Geographically Weighted Poisson Regression (Regressão de Poisson Geograficamente Ponderada) –, e o modelo GWPR semi-paramétrico. O modelo que revelou um melhor ajuste aos dados foi o GWPR com oito variáveis explicativas: Poder de compra per capita; Taxa de população jovem residente; Taxa de retenção e desistência no ensino básico; Taxa bruta de escolarização; Número médio de alojamentos de familiares clássicos; Taxa de benificiários do RMG e RSI; Taxa de desempregados; Proporção de população estrangeira. Em particular, o modelo apresenta melhor ajuste nos municípios da área metropolitana de Lisboa e do Porto. Os resultados mostram que a magnitude da associação de cada uma das variáveis independentes com os crimes contra o património varia no território. Com este estudo, espera-se contribuir para uma melhor compreensão do fenómeno da criminalidade em Portugal, na medida em que os resultados permitem identificar quais são os fatores demográficos e socioeconómicos mais relevantes em cada município.Criminality is one of the most problematic social phenomena that has been gaining special relevance since the late twentieth century. It is a dynamical phenomenon distributed according to spatial trends. Over the course of time, several research fields have focused on the study of this subject, in order to unravel innovative ways of examining the information, making it possible to better understand the factors associated with crime, and find solutions to prevent and address this issue. The subject of criminality, like many others, is strongly related with territorial distribution and therefore, it may be well characterized by a geographical assessment. Thus, crime monitoring through geographic information analysis tools stands as an important contribution to the study of crime. This study explores the spatial distribution of crime against property in mainland Portugal, with the primary goal of investigating which demographic and socioeconomic factors may be associated with this problem in each municipality. For this purpose, Geographic Information System (GIS) tools, exploratory analysis techniques, descriptive statistics, and spatial modelling have been applied. An initial exploratory analysis identified the existence of autocorrelation and spatial heterogeneity in the rate of crimes against property, thus supporting the use of spatial regression models. Hence, Poisson regression models were applied and compared, namely the global model, the local GWPR model – Geographically Weighted Poisson Regression –, and the semi-parametric GWPR model. The model that revealed a better fit to the data was the GWPR model, with eight explanatory variables: Purchasing power per capita; Rate of young resident population; Retention and dropout rates in basic education; Gross school enrollment rate; Average number of households in a classical family; Rate of RMG and RSI beneficiaries; Unemployment rate; Ratio of foreign population. The model presents a better fit in the municipalities on the metropolitan areas of Lisbon and Porto. The results show that the correlation degree between each of the independent variables and the crimes against property varies across the territory. The aim of this investigation is to potentially improve the understanding of the crime problem in mainland Portugal, given that the results allow to identify the most relevant demographic and socioeconomic factors in each municipality
|