makeneighborsw: Spatial weight matrix

In tibo31/GeoXp: Interactive Exploratory Spatial Data Analysis

Spatial weight matrix

Description

The function makeneighborsw() create a spatial weight matrix based on a given number of nearest neighbors (option "neighbor" by default), based on a threshold distance (option method="distance") or both these 2 methods.

Usage

makeneighborsw(coords, method = "neighbor", m = 1, d, cum = TRUE)

Arguments

coords	a matrix of spatial coordinates
method	"neighbor" by default, "distance" or "both"
m	number of nearest neighbors
d	threshold point
cum	if cum=TRUE, W is the sum of spatial weight matrix based on k nearest neighbours (for k \leq m; if FALSE W is the spatial weight matrix based only on m^{th} nearest neighbours

Details

In the case of method="neighbor", for each site, we order the other sites by their distance from this site. If cum=TRUE, for i, if j is among the m^{th} nearest sites, then :

W_{ij}=1

else

W_{ij}=0

If cum=FALSE, for

s_i

, if

s_j

is the m^{th} nearest site, then :

W_{ij}=1

else

W_{ij}=0

In case of ties, the nearest neighbour is randomly chosen.

In the case of method="distance", if site i is seperated from j by a distance lower or equal to a given threshold :

W_{ij}=1

else

W_{ij}=0

In the case of method="both" W must verify the two first conditions.

Value

A spatial weight matrix of size n \times n

Note

This function is not optimised for large dataset. User could find similar functions in the package spdep (dnearneigh and knearneigh). However, these functions don't offer the possibility to use the two criteria in the same time. Moreover, an inconvenient of makeneighborsw is that the result is included in a matrix object whereas most of functions of GeoXp use the nb structure for spatial weight matrix. An issue is to use the mat2listw function and then selecting the nb part, like the in the examples.

Author(s)

Aragon Y., Thomas-Agnan C., Ruiz-Gazen A., Laurent T., Robidou L.

References

Thibault Laurent, Anne Ruiz-Gazen, Christine Thomas-Agnan (2012), GeoXp: An R Package for Exploratory Spatial Data Analysis. Journal of Statistical Software, 47(2), 1-23.

Roger S.Bivand, Edzer J.Pebesma, Virgilio Gomez-Rubio (2009), Applied Spatial Data Analysis with R, Springer.

See Also

moranplotmap,normw

Examples

##
# data auckland
auckland <- sf::st_read(system.file("shapes/auckland.shp", package="spData")[1])

coords <- cbind(auckland$Easting[1:10], auckland$Northing[1:10])

# matrix based on 5 nearest neighbors
W <- makeneighborsw(coords, method = "neighbor", m = 3)

# matrix based on a threshold distance
W1 <- makeneighborsw(coords, method = "distance", d = 20)

# matrix based on the two methods
W2 <- makeneighborsw(coords, method = "both", m = 3, d = 20)

# representation of the 3 spatial weight matrices
op <- par(mfrow = c(2, 2))
plot(mat2listw(W), coords, col = "lightblue1", main = "neighbor")
plot(mat2listw(W1), coords, col = "lightblue2", main = "distance")
plot(mat2listw(W2), coords, col = "lightblue3", main = "both")
par(op)

tibo31/GeoXp documentation built on April 8, 2023, 7:50 a.m.