dnearneigh

dnearneigh is a function in the R package spdep used to construct a neighbourhood (neighbor) structure based on pairwise distances among a set of points. It returns a neighbour list object that identifies, for each observation, which other observations lie within a specified distance band. This type of structure is commonly used to build spatial weights for analyses such as Moran’s I, spatial regression, and other methods in spatial econometrics and geography.

How it works

The function takes coordinates or a spatial object and two distance thresholds, d1 and d2. For each observation i, all other observations j are considered neighbors of i if the distance between i and j satisfies d1 <= distance(i, j) <= d2. Distances can be computed as Euclidean distances by default, or as great-circle distances if longlat = TRUE and the coordinates are geographic (longitude and latitude). The diag option controls whether an observation is allowed to be its own neighbor; typically this is FALSE unless d1 <= 0 <= d2. The row.names argument can supply names for the observations.

Output and usage

The result is an object of class nb, a list where each element contains the indices of neighboring observations for the corresponding location. This nb object can be converted to a weights object (for example, via nb2listw) and used in spatial statistical routines. dnearneigh is commonly contrasted with other neighbor-building functions like knearneigh (fixed k neighbors) or tri2nb (triangulation-based neighbors).

Example (conceptual)

coords <- cbind(x coordinates, y coordinates)

nb <- dnearneigh(coords, 0, 1000) # neighbors within 0 to 1000 units