sfdep (version 0.2.5)

local_colocation: Local indicator of Colocation Quotient

Description

The local indicator of the colocation quotient (LCLQ) is a Local Indicator of Spatial Association (LISA) that evaluates if a given observation's subcategory in A is colocated with subcategories in B. Like the CLQ, the LCLQ provides insight into the asymmetric relationships between subcategories of A and B (where B can also equal A) but at the local level.

The LCLQ is defined using Gaussian kernel weights and an adaptive bandwidth (see st_kernel_weights()). However, any type of weights list can be used. Kernel weights are used to introduce a decay into the calculation of the CLQ. This ensures that points nearer to the focal point have more influence than those that are more distant.

Usage

local_colocation(A, B, nb, wt, nsim)

Value

a data frame with as many rows as observations in A and two times as many columns as unique values in B. Columns contain each unique value of B as well as the simulated p-value for each value of B.

Arguments

A

a character or factor vector.

B

a character or factor vector.

nb

a neighbors list e.g. created by st_knn() or st_contiguity()

wt

a weights list. Recommended that it is a Gaussian kernel weights list using an adaptive bandwidth e.g. created by st_kernel_weights(nb, geometry, "gaussian", addaptive = TRUE) that does not include the self.

nsim

default 99. An integer representing how many simulations to run for calculating the simulated p-values.

Details

The LCLQ is defined as \(LCLQ_{A_i \to B} = \frac{N_{A_i \to B}}{N_B / (N - 1)}\) where \(N_{A_i \to B} = \sum_{j = 1(j \ne i)}^{N}(\frac{w_{ij}f_{ij}}{\sum_{j = 1(j \ne i)}^{N}w_{ij}})\). And the weights matrix, wij, uses adaptive bandwidth Gaussian kernel weights.

LCLQ is only calculated for those subcategories which are present in the neighbor list. If a subcategory is not present, then the resultant LCLQ and simulated p-value will be NA.

References

Fahui Wang, Yujie Hu, Shuai Wang & Xiaojuan Li (2017) Local Indicator of Colocation Quotient with a Statistical Significance Test: Examining Spatial Association of Crime and Facilities, The Professional Geographer, 69:1, 22-31, tools:::Rd_expr_doi("10.1080/00330124.2016.1157498")

Examples

A <- guerry$main_city
B <- guerry$region
geo <- sf::st_centroid(sf::st_geometry(guerry))
nb <- include_self(st_knn(geo, 5))
wt <- st_kernel_weights(nb, geo, "gaussian", adaptive = TRUE)
res <- local_colocation(A, B, nb, wt, 9)
tail(res)