주류업소밀집도와 성범죄 간의 상관관계에 대한 공간가중 포아송 회귀분석

This study aimed to examine the relationship between alcohol outlet density and sex crimes in the census output areas of Y-gu, X city, South Korea. This study observed sex crimes, as a count variable, demonstrating a zero-inflated and over-dispersed distribution. Zero-inflated negative binomial regression(ZINBR) or negative binomial regression(NBR) models are usually used to analyze a zero-inflated and/or over-dispersed dataset. However, these models might be less suitable when spatial dependence is observed among variables. Thus, identifying the existence of spatial dependence among structural variables, this study applied geographically weighted Poisson regression(GWPR) model on analyzing the relationship between sex crimes and alcohol outlet density. The GWPR model produced the outcome in both global and local perspective. In the global perspective, this study found that in comparison with the ZINBR and the NBR model, the GWPR model fit better with lower mean absolute deviation and mean squared error. Also, in the local perspective, this study identified that the relationship between the alcohol outlet density and sex crimes within each census output area significantly varied.