——————————————————————–
While Euclidean buffers serve as a good starting point, the weaknesses compared to using walksheds is apparent. Measuring distance to a service “as the crow flies” likely overstated the greenway accessibility of many neighborhoods. Note: this diagram is somewhat exaggerate to illustrate the constraint of Euclidean buffers. A service point was placed in an relatively inaccessible location to make the walkshed smaller.
—————————————————————————–
A major limitation with this study is the use of geodesic centroids to measure proximity of a neighborhood to a greenway. Reality shows that a population weighted centroid will often be farther or closer to the service. In this case, a tract that falls inside the catchment area likely would not have if population weighted centroids been used.
While Euclidean buffers serve as a good starting point, the weaknesses compared to using walksheds is apparent. Measuring distance to a service “as the crow flies” likely overstated the greenway accessibility of many neighborhoods. Note: this diagram is somewhat exaggerate to illustrate the constraint of Euclidean buffers. A service point was placed in an relatively inaccessible location to make the walkshed smaller.
In this case, the entire census tract is given the same greenway accessibility ratio even though it is divided by a major highway with few crossing points. If walksheds and sidewalk coverage were considered, it would likely represent the dynamic between greenways and residents better.
——————————————————————–
Spatial Distribution of Negative Binomial Regression Model Residuals:
White tracts indicate near-perfect predictions. The lack of autocorrelation indicates the model is not systemically skewing predictions. In lieu of sub-tract level population data, this model uses percent coverage per tract by parks and greenways to calculate supply. This helps resolve many of the inaccuracies produced by using geodesic centroids in the previous model.
White tracts indicate near-perfect predictions. The lack of autocorrelation indicates the model is not systemically skewing predictions. In lieu of sub-tract level population data, this model uses percent coverage per tract by parks and greenways to calculate supply. This helps resolve many of the inaccuracies produced by using geodesic centroids in the previous model.
Analysis Upgrade: Walksheds.
Swapping Euclidean point-to-point measurements strengthens the validity of the findings. Although there is still room for improvement in a model like this. Park/greenway quality, sidewalk quality, crime rate, noise pollution, and nonlinear thresholds (ex: diminishing returns after 0.1 miles) were not accounted for
Swapping Euclidean point-to-point measurements strengthens the validity of the findings. Although there is still room for improvement in a model like this. Park/greenway quality, sidewalk quality, crime rate, noise pollution, and nonlinear thresholds (ex: diminishing returns after 0.1 miles) were not accounted for
Our direct model identifies socioeconomic status (Income and Race) as the primary drivers of obesity. While ‘Zero Vehicle Access’ and ‘Park Proximity’ trend in the expected directions, their impact is statistically secondary to systemic economic factors.
Key Insight: Further mediation analysis revealed that Park Access becomes a significant protective factor only when modeled through the pathway of safe pedestrian (sidewalk) infrastructure.
Key Insight: Further mediation analysis revealed that Park Access becomes a significant protective factor only when modeled through the pathway of safe pedestrian (sidewalk) infrastructure.