After exploring my dataset for Ph.D. thesis and learning several spatial econometric techniques, I successfully applied ordinary least squares (OLS), logistic regression, Spatial Autoregressive models [i.e., Spatial Lag model(SLM), Spatial Error Model(SEM), Spatial Durbin Model(SDM)], and most importantly Geographically Weighted Regression (GWR), and Geographically Weighted Logistic Regression models to find evidence of spatial and socioeconomic inequality in flood risk. The performance of all regression models was significantly improved when I accounted for spatial heterogeneity at the local level compared to non-spatial global models such as OLS and logistic regression.
I am amazed that several research papers were published so far in high-rank journals based on global regression results only, which I could have done a couple of months ago. The results do not make sense because the nature of the spatial heterogeneity could prevail in flood exposure. In my view, flood exposure and/ effects of flood risk cannot be locally independent by census tracts or dissemination areas or census subdivisions; they must be spatially autocorrelated. There remain ripple effects, spillover effects or indirect effects to adjacent neighbourhoods and to the overall economy. Populations from affected or flooded neighbourhoods could move to nearby safer neighbourhoods, looking for jobs and safe accommodation. Many other indirect socio-demographic effects could prevail around the flooded neighbourhoods. Do you agree? Please, justify your response.