add_spatial_lags: Add Spatial Lags of a Variable to a Dataset

In ercbk/ebtools: Eric's Data Science Toolbox

Add Spatial Lags of a Variable to a Dataset

Description

Computes spatial lags for a given numeric variable in a dataset using a neighborhood list from the spdep package. It supports inverse distance weighting, exponential, and double power decay weighting methods along with various normalization procedures.

Usage

add_spatial_lags(nblist, y, .data, lags, type = NULL, parallel = FALSE, ...)

Arguments

nblist	An object of class "nb" from the spdep package which represents the neighborhood structure.
y	A character string indicating the name of the numeric variable in .data for which spatial lags will be computed.
.data	A data frame of class "sf" containing the variable specified by y.
lags	A numeric value specifying the number of spatial lags to compute.
type	A character string indicating the type of spatial weights to use. Accepted values are "idw" (inverse distance weighting), "exp" (exponential), "dpd" (double power decay), or NULL (default), which applies standard adjacency-based weighting.
parallel	(default: FALSE) Logical indicating whether to use parallel processing. Requires having purrr (>= 1.0.4.9000) installed, mirai (>= 2.1.0.9000) package installed and loaded, and setting mirai::daemons() to the number of desired processes. (See Examples)
...	Additional arguments passed to spdep::nb2listw() or spdep::nb2listwdist(), such as the style argument specifying the normalization method or zero-policy which indicates whether or not to permit the weights list to be formed with zero-length weights vectors.

Details

To obtain a neighborhood list of class "nb," first, a neighborhood algorithm is fit to the geometry of sf dataset. Then, that object is coerced into a neighborhood list. For a workflow, see the Misc section in the Geospatial, Spatial Weights note of my Data Science notebook.

• Valid Types: ("idw", "exp", "dpd").

• Valid Styles: ("W", "B", "C", "S", "U", "minmax", "raw").

• See the Spatial Weights section in the Geospatial, Spatial Weights note of my Data Science notebook for details

The spatial weights summary is extracted from the output of printing the spdep::nb2listw or spdep::nb2listwdist object. It contains characteristics such as the number of regions, number of nonzero links, percentage of nonzero weights, average number of links.

• n: This refers to the number of regions (or spatial units) in your dataset.

• nn: This refers to the total number of possible pairwise relationships between the regions. It is calculated as n × n. This represents the total number of possible links if every region were connected to every other region, including itself.

• S0: This is the sum of all weights.

• S1: This is related to the sum of the squares of the weights.

• S2: This is related to the sum of the products of the weights for each pair of neighbors.

• S0, S1, S2 are constants used in inference for global spatial autocorrelation statistics

Value

A tibble containing the original dataset with additional columns for each computed spatial lag. The spatial lag columns are named "spatlag_<lag>_<y>". The output also includes weight summary attributes named "summ_wgts_spatlag_<lag>".

Examples

library(spdep, quietly = TRUE)

ny8_sf <-
  st_read(system.file(
    "shapes/NY8_bna_utm18.gpkg",
    package = "spData"),
    quiet = TRUE)

dplyr::glimpse(ny8_sf)

ny8_ct_sf <-
  st_centroid(st_geometry(ny8_sf),
              of_largest_polygon = TRUE)


ny88_nb_sf <-
  knn2nb(knearneigh(ny8_ct_sf,
                    k = 4))

# Compute spatial lags
tib_spat_lags <-
  add_spatial_lags(
    nblist = ny88_nb_sf,
    y = "PCTOWNHOME",
    .data = ny8_sf,
    lags = 2,
    type = "dpd",
    dmax = 25000,
    style = "W",
    zero.policy = TRUE
  )

tib_spat_lags |>
  dplyr::select(PCTOWNHOME,
                spatlag_1_PCTOWNHOME,
                spatlag_2_PCTOWNHOME) |>
  dplyr::glimpse()

cat(attributes(tib_spat_lags)$summ_wgts_spatlag_1, sep = "\n")

rlang::check_installed(
  "mirai (>= 2.1.0.9000)",
  action = function(...) {
    remotes::install_version('mirai',
                             version = ">= 2.1.0.9000",
                             repos = c('https://shikokuchuo.r-universe.dev',
                                       'https://cloud.r-project.org'))
  }
)

library(mirai)

daemons(2)

tib_spat_lags_para <-
  add_spatial_lags(
    nblist = ny88_nb_sf,
    y = "PCTOWNHOME",
    .data = ny8_sf,
    lags = 2,
    type = "exp",
    zero.policy = TRUE,
    parallel = TRUE
  )

daemons(0)

ercbk/ebtools documentation built on Feb. 22, 2025, 1:51 p.m.