Table 3. Protocol for identifying and managing point-in-polygon aggregation uncertainty.

Stage	Procedures
1) Observation	Overlay point data (e.g., incidents) with polygon data to which they will be aggregated (e.g., census tracts). Observe the distribution of points in relation to polygons–are they visibly within the polygons, or do they appear to also intersect with boundaries? If visual observation alone can confirm absence of intersection, then the study may proceed without using additional stages in the protocol. If this cannot be confirmed, then the next stage of analysis should be performed.
2) Analysis	a) Conduct Near Analysis to calculate distances between points. If results confirm the absence of intersection, then the study may proceed without using additional stages in the protocol. If this cannot be confirmed, then secondary and/or tertiary analyses are required. b) If points intersect polygons boundaries (NEAR_DIST = 0), then query the attribute table with NEAR_DIST calculations to identify the points intersecting the boundaries. This will identify where uncertainty will arise in the aggregation process. If working with a large dataset, it may also be useful to look for clustering in the data where this concern may be more prominent. c) Conduct Kernel Density Estimation (KDE) to visualize hotspots of potential uncertainty; use of spatial statistics such as Local Moran’s I or GI* may be used to further quantify spatial autocorrelation in these areas of concern.
3) Management	If intersecting points are numerous and widespread such that they cannot be studied and then assigned to an appropriate polygon on a case-by case basis, then use the polygon-in-point (Option 4) spatial join approach to ensure that these points are not counted multiple times.