2-D kernel density estimate of geocoded incidents

Evaluates a Gaussian KDE on incident lat/long and returns summary statistics plus the (lat, lon) of the densest observation. Prefers MASS::kde2d when available; otherwise uses a pure base-R Gaussian kernel evaluated at the observation points (i.e. kernel density at each datum).

Usage

morie_tps_kde_density(
  df,
  bandwidth = 0.005,
  ds_name = "?",
  lat_col = "LAT_WGS84",
  lon_col = "LONG_WGS84"
)

Arguments

df

Incident-level data.frame.

bandwidth

Bandwidth multiplier passed to the 2-D KDE; see the h argument of the native morie_kde2d.

ds_name

Tag for the result title.

lat_col, lon_col

WGS84 column names.

Value

A named list with summary stats including max/mean/median density and the (lat, lon) of the densest observation.

Examples

set.seed(2026)
df <- data.frame(
  LAT_WGS84 = 43.6 + rnorm(120, 0, 0.05),
  LONG_WGS84 = -79.4 + rnorm(120, 0, 0.05)
)
morie_tps_kde_density(df, bandwidth = 0.01)
#> KDE -- ?
#> ========
#> Call: morie_tps_kde_density(df=<120r>, bandwidth=0.01) 
#> 
#>   Geocoded incidents      120
#>   Bandwidth               0.01
#>   Max density (at obs)    537.06
#>   Mean density            139.44
#>   Median density          129.81
#>   Lat at max-density obs  43.563
#>   Lon at max-density obs  -79.427
#>   Backend                 morie_kde2d
#> 
#> Across 120 geocoded incident(s), peak kernel density is 537.0641 (mean 139.4419, median 129.8126; backend: morie_kde2d). The densest cluster centre is at (lat=43.5632, lon=-79.4272). Smaller bandwidth exposes finer local clusters; larger bandwidth oversmooths.