A1305
Title: Modelling gunfire in Washington, D.C. using a spatiotemporal Hawkes process with nonseparable triggering function
Authors: Tom Stindl - UNSW (Australia) [presenting]
Jeffrey Kwan - University of New South Wales (UNSW) Sydney (Australia)
Feng Chen - UNSW Syd (Australia)
Yongtao Guan - Chinese University of Hong Kong, Shenzhen (China)
Abstract: The United States has an epidemic of gun violence, with the risk of firearm-related death being significantly higher compared to other high-income nations. Implementing intervention strategies to reduce firearm-related activity requires an improved understanding of the temporal, spatial, and contagious dynamics of gunfire. A spatiotemporal Hawkes process is introduced to model the space-time clustering of gunfire in Washington, D.C., in 2018 using data obtained from the ShotSpotter technology. The model has three essential features. First, space-time interactions are introduced for contagious gunfire by modeling the triggering function as a mixture of space-time functions. Second, the productivity of contagious gunfire varies across both time and space. Third, the intensity of non-contagious gunfire incorporates demographic and socioeconomic variables at the census block level to better understand the characteristics that lead to persistent gunfire in particular communities. The intensity also includes cyclic trends such as daily, weekly, and long-term effects. An estimation procedure is proposed within the expectation-maximization framework, and the results are discussed and interpreted.
