The collaborative mobile app Fogo Cruzado delivers instant alerts every time a user reports a gun fire in the Rio de Janeiro metropolitan area (Brazil). This app contributes to public safety and generates a valuable dataset comprising the location and time of occurrence of gun shootings whose analysis may allow reserchers to understand gun fire dynamics in Rio and support the development of crime reduction plans. Prior to applying existing crime forecasting methods, such as kernel hotspot maps and self-exciting, we should test if the gun fire patterns meet their assumptions. For this purpose, we have applied nonparametric first and second-order point process inference. The kernel intensity estimator describes the spatial distribution of gunfire and identifes chronic hotspots. The nonparametric test for comparison of first-order intensities found differences between gun fires with and without fatalities or police intervention. The recently developed log-ratio based first-order separability test found that the spatial distribution of gun fire, fatalities and police presence varied over time. Finally, spatiotemporal inhomogeneous K-tests detected clustering between gun fire events, fatalities and police interventions. These results suggest that we could consider a self-exciting point process with nonseparable background component as a starting point in the development of a suitable approach to forecast gun fire hotspots in Rio de Janeiro.