Analysing Shooting Incidents in New York Using Spatial Regression Techniques

Spatial regression techniques usually applied to area data are here adapted to work with network data - in this case road sections. An example is given exloring spatial patterns in shooting incidents in a neighbourhood of New York City, making use of open data. Risks of a shooting occuring are modeled using both Poisson and negative binomial distributions, and compared using an Aikake Information Criterion (AIC) approach - which suggests that the negative binomial model (in which events tend to cluster) is more plausible than the Poisson.