ABSTRACT
Distance-based spatial weights define a neighbor condition based on the distance between spatial units. This is particularly relevant for observations represented as points, but it can be extended to the distance between central points or centroids of polygon data. Different distance metrics are reviewed, including Euclidean distance, Manhattan block distance and the Minkowski metric. When coordinates are not Cartesian, the appropriate distance is great circle distance or arc distance. In addition, a general distance metric can be computed between observations as points in multivariate attribute space.
Two main types of distance-based weights are implemented in GeoDa's Weights Manager, either based on a distance band or on a k-nearest neighbor relationship. As for contiguity-based weights, characteristics can be obtained in the form of summary statistics, a connectivity histogram or map and a connectivity graph. Special cases include block weights and space-time weights. Operations on spatial weights, such as intersection, union and make symmetric are illustrated as well.
