Table 7 Definition, advantages and disadvantages of the spatial weight matrix.
Matrix type | Definition | Advantages | Disadvantages |
|---|---|---|---|
Queen adjacency | Two spatial units are considered neighbors if they share any length of common boundary (edge) or vertex | Simple and intuitive, suitable for most polygonal data | May result in many cells having numerous neighbors (e.g., large polygons) or some cells having no neighbors (island problem) |
Rook adjacency | Two spatial units are considered neighbors if they share a common edge (not just a vertex) | This is stricter than the Queen definition, excluding cases connected solely by vertices | Uneven neighbor counts and island problems may still occur |
Threshold distance | Set a threshold distance (d). If the distance between the centers of mass of two units is less than or equal to this distance, they are considered neighbors | Highly flexible, suitable for irregular units; ensures all units have at least some neighbors | The choice of threshold distance d is subjective; it may result in global connectivity (all cells are neighbors) or local connectivity imbalance |
K-Nearest neighbors | Identifies the K nearest neighboring cells for each spatial cell | Effectively addresses uneven neighbor counts by guaranteeing each cell has the same number of neighbors; particularly suitable for datasets with highly variable cell sizes | May result in asymmetric neighbor relationships (i is a neighbor of j, but j is not necessarily a neighbor of i), though GeoDa typically symmetrizes these relationships |
