Spatial statistical modeling of the pollution impact of old industrial sites on colon and lung cancer incidents in New York State, USA

This research visualizes the spatial patterns of diagnosed colon and lung cancer mortalities across the New York State. Kernel density analysis was applied to visualize the spatial patterns of old industrial sites across the state. Geographically Weighted Regression (GWR) was applied to model the possible pollution impact of old industrial sites on colon and lung cancer incidents. GWR is a local spatial statistical model analyzing the relations of non-stationary factors across the space. It generates a separate spatially-weighted least square regression for each of the observations across the study area. The results shows that the high density clusters of the old industrial sites occur in the large urban areas of New York City, Rochester, Buffalo, Syracuse, and Albany. GWR modeling indicates that high local R2 coincides with moderate cancer mortality incidents. The overall GWR predictions are satisfactory as the R2 of colon cancer is 0.8848 and that of lung cancer is 0.8755. However, AICc values are large for both types of cancer predictions.