kernel_density: Kernel density estimation with plot functions

In gmcirco/crimtools:

Kernel density estimation with plot functions

Description

Computes a kernel density estimate for an 'sf' point pattern.

Usage

kernel_density(
  x,
  region,
  bdw = "auto",
  bdw.fun = NULL,
  kernel = "gaussian",
  npixel = 100,
  plot = T,
  return_grid = F
)

Arguments

x	an 'sf' point pattern. Should be the point pattern of interest.
region	an 'sf' polygon feature that defines the study region.
bdw	Bandwidth for kernel smoothing. Should be in projected values (meters or feet). If no values are specified, default behavior is 'auto' which will apply a bandwidth estimation function specified in the argument 'bdw.fun'.
bdw.fun	Function to be used to automatically estimate a bandwidth. Defaults to'spatstat.core::bw.ppl', but any function that returns a single numeric value can be substituted. Many of the built-in 'spatstat' functions work well, such as 'bw.diggle'.
kernel	Type of kernel to be used for the KDE estimation. Default is 'gaussian' but can also use 'epanechnikov', 'quartic', or 'disc'.
npixel	Dimensions of grid points to be used. Should be specified in projected values (meters or feet).
plot	Should a quick plot be generated?
return_grid	Should the raster grid be returned? If TRUE, returns a SpatialPolygonsDataFrame, otherwise returns a dataframe with columns for the density value, X, and Y. Defaults to FALSE.

Details

The kernel_density function uses functionality in the 'spatstat' package to compute kernel density estimates for 'sf' objects. The function can be used to return a dataframe with density estimates for grid cells. This can further be used for visualization purposes in 'ggplot2'. This function can also return the raw 'sp' object as a 'SpatialPolygonsDataFrame'. This can easily be converted to an 'sf' object, or exported as a shapefile for use in other programs.

Examples

data("newhaven")
data("nh_hom")

# kernel density estimation with default bandwidth
kde_out <-
    kernel_density(
        x = nh_hom,
        region = newhaven)

head(kde_out)

# Plotting kde in ggplot
library(ggplot2)

ggplot(kde_out) +
 geom_tile(aes(x = X, y = Y, fill = density)) +
 coord_equal() +
 scale_fill_viridis_c() +
 theme_void()

# kernel density estimation with custom bandwidth
# and quartic kernel

kde_out_quart <-
   kernel_density(
       x = nh_hom,
       region = newhaven,
       kernel = "quartic",
       bdw = 1500)

gmcirco/crimtools documentation built on May 22, 2022, 1:31 a.m.