Response to “Crime Places in Context”
- 232 Downloads
The recent paper “Crime Places in Context: An Illustration of the Multilevel Nature of Hot Spot Development” (Deryol et al. 2016), published in Journal of Quantitative Criminology, uses multilevel Poisson regression analysis to evaluate factors which contribute to local crime rates. In particular, the authors test three hypotheses concerning the three-way interaction between nearby carry-out liquor stores, on-premises drinking establishments, and bus routes, to determine whether “it is a combination of risky nodes and paths that are more important than any one single risky facility”. However, the analysis contains statistical flaws which render it unable to test the hypotheses and invalidate the conclusions presented. Most importantly, their interaction term is not an interaction at all.