The nearest neighbor index tells us whether values are clustered or randomly dispersed. The nearest neighbor index measures the spatial distribution from 0 (clustered) to 1 (random) to 2.15 (regular). The nearest neighbor index compares the differences between nearest points and distances that would be expected based on chance. The mean nearest neighbor distance was 217 feet while the mean nearest neighbor distance under randomness was 499 feet. The nearest neighbor index, the ratio of the actual nearest neighbor distance to the random nearest neighbor distance, is 0.4352. The Z-value of -46.4490 is highly significant with a p-value of 0.0001. This means that crimes are not randomly dispersed, and instead the distribution of nearest neighbors of violent crimes in St. Louis City is much smaller than 1, meaning that values are clustered. Additionally, the graph above visualizes the nearest-neighbor index values over 25 orders. The values only reach above 1 until 21 orders, further strengthening the idea that violent crimes are clustered within the study area.
It’s expected that crimes occur within the same area, as criminal pattern theory suggests that criminal offenders area influenced by their daily activities and routines, and they tend to concentrate in areas that are known to them. Broken window theory also explains why crimes may be clustered. Broken window theory suggests that any visible signs of crime and disturbance will generate more crimes.
Moran’s I
Moran’s I can provide information about the scale of spatial auto-correlation, and whether these values are concentrated near each other or diffuse moving away from a point. For the Moran’s I, violent crimes were aggregated into each census tract to see if the incidents are associated with the distribution of the census tracts. However, the results from the Moran’s I however presented the p-value as not significant. This likely means that the census tracts are too big of an area and the points were too far apart from each other. Information on census blocks was not available for crimes, however this would solve the problem in calculating a significant Moran’s I values.
Violent Crime in the City of St. Louis 