2.6. Sensitivity analyses

For sensitivity analyses, we estimated the Community Severance Index using principal component pursuit combined with factor analysis adding as an input a variable representing the ratio of the area of sidewalks (Rhoads et al., 2021) over the area of roads (NYC Open Data, 2022) within walkable distance (0.5 miles) of each CBG centroid. We also included two datasets related to crossings from NYC Open Data (NYC Open Data, 2022): enhanced crossings and accessible pedestrian signals in NYC. Enhanced crossings are intended to provide a safe place for crossing the street when there is no traffic signal. Accessible pedestrian signals are installed to ease crossing the street to visually impaired people or those who have low vision. These variables are excluded from the main analysis as both sidewalk and crossing data are not available nationwide.

We also reran the models linking Community Severance Index and vehicle collisions additionally adjusting for the National Walkability Index per CBG. We also reran the models additionally adjusting for the annual average daily traffic from ESRI per CBG. We repeated analyses using the vehicle collisions data restricting to instances with pedestrians or cyclist involved (i.e., injured or killed). Lastly, we used a flexible function of space that adjusts for spatial correlation among CBGs to assess the robustness of our results to spatial dependence in vehicle collisions. We added a two-dimensional term for the coordinates (latitude and longitude) of each CBG centroid using a tensor product in the model.

All statistical analyses were performed using the R Statistical Software, version 4.2.2 (R Core Team, 2022). All data and code are available at https://github.com/jaime-benavides/community_severance_nyc.