These two scatter plots are standardized representations showing the correlation between the dependent variable (SUM_ALLOC: total number of housing assistance per grid cell) and the independent variables (MIN_NEW: minimal distance from the center of grid cell to the center of polluted area, DIST_MINSE: minimal distance from the center of grid cell to the closest service station).
The purple line is the regression line and shows in this case a negative autocorrelation between the values (slope b = -0.235 and -0.221).
Discussion
As indicated in the results section, both scatterplots in figures 2 and 3 show a negative autocorrelation between the housing assistance and the distances. This tends to confirm the original hypothesis stating that more vulnerable people are located closer to service stations and their related polluted areas. This class of people seem therefore more prone to pollution, as some housings are located within a 75 m radius from service stations mentioned in the paper \citep{Morales_Terr_s_2010} as the region of direct pollution impact due to the stations emissions.
This comes as an interesting result and adds up to the fact that this group of population usually resides closer to road networks. Aggravated health risk factors could then be considered.
Nevertheless, this conclusion is to be tempered. Indeed, the determination coefficient (R2) is close to zero in both cases : 0.0533 and 0.0488. This indicates the regression line does not fit the data very well and high residues are to be expected.
This bad fitting is emphasized by the vulnerability grids obtained when applying the non-spatially and spatially weighted regression. The results are presented in the following figures.