Kilian: DISCUSSION
First of all, in our discussion we have to be critical with the methodology we used: two of the datasets (neo_mort and vio_tot) were downscaled from a province level to a district level. This procedure introduces a bias in the linear regression as for the two variables above we omit the variation among the districts of the same province. Thus, where the other variables have a large variation among the same province, the residuals are high as that variation cannot be explained by the regression. However, we can still explain some of the results that we obtained.
The