Project Overview and Aims

A fundamental challenge in the social, physical, environmental and health sciences is the identification of spatial clusters.  Spatial clusters are locations that exhibit elevated levels of resources (e.g., mineral deposits), disease (e.g., Ebola), crime, or vulnerability to extreme events (e.g., flooding).  Many of the existing statistical approaches used for identifying spatial clusters are flawed because they use pre-defined geometric shapes (e.g., circles or ellipses) to identify the clusters.  The problem with these approaches is that spatial clusters rarely assume a pre-defined shape.  Instead, many clusters display elongated or irregular shapes because of how or where they occur.    

The first aim of this project was to develop more rigorous and efficient spatial cluster detection approaches that do not rely upon geometric scanning windows.  These techniques were designed to minimize error, improve accuracy and be applied over a wide range of spatial scales and research domains.    

The second aim of this project was to apply the developed spatial clustering methods to explore the connection between alcohol outlets and urban violence.  Although research suggests that alcohol outlets contribute to violent outcomes in cities, little is known about how clusters of alcohol outlets may influence offending, victimization and injury.  The research examined alcohol outlets and violence, focusing on the cities of Philadelphia, Pennsylvania, Cincinnati, Ohio and Seattle, Washington. 

Results

Several new statistical approaches were developed over the lifespan of this project, including the CM-LLR model (Murray et al., 2014).  Results suggest that the CM-LLR model provides significant improvements in cluster detection power, on the order of 20% - 72%, when compared to the traditional spatial clustering approaches that rely upon geometric windows.  The accuracy of the CM-LLR approach is also extremely good.  The CM-LLR did not misidentify any spatial units during empirical testing on assaults in the city of Cincinnati for 2010 (Grubesic et al., 2015).  The CM-LLR is also both fast and efficient, displaying average solution times of less than 1 second.  This means it can provide almost instantaneous results, a feature that is important when monitoring and tracking dynamic spatial clusters or event outbreaks (Grubesic et al., 2015). 

One of the most compelling results concerning alcohol outlets and their spatial distribution was uncovered in the city of Philadelphia.  Pennsylvania is currently considering legislation to privatize alcohol sales.  Thus, rather than forcing consumers to purchase beer, wine and spirits at state sanctioned stores and distributors, alcohol would instead be sold by major retail chains (e.g. drug stores, grocery stores, etc.), corner stores and any retailer that could obtain a license.  However, the state of Pennsylvania also maintains a quota law for outlets and suite of geographic proximity restrictions that control the number and location of outlets licensed to sell alcohol.  For example, only one retail license can be issued for every 3,000 residents in each county.  Further, alcohol outlets cannot be within 200 feet of each other and must be at least 300 feet away from sensitive facilities such as schools, churches, hospitals, parks and playgrounds.  At question, then, is determining what the spatial distribution of alcohol outlets would look like if privatization occurred in Pennsylvania and the proximity restrictions remained in place. 

Findings suggest that current state policies on alco...

Please report errors in award information by writing to: awardsearch@nsf.gov.