hotspot_dual_kde() produces invalid results with lat/lon data #67
Description
library(sfhotspot)
# Running dual KDE on lat/lon data produces missing KDE estimates
hotspot_dual_kde(memphis_robberies, memphis_population) |> summary()
#> Cell size set to 0.00524 degrees automatically
#> Bandwidth set automatically based on rule of thumb.
#> ℹ Bandwidth for `x` = 0.06128371 degrees.
#> Bandwidth set automatically based on rule of thumb.
#> ℹ Bandwidth for `y` = 0.05693893 degrees.
#> Data transformed to "WGS 84 / UTM zone 16N" co-ordinate system.
#> ℹ CRS code: "EPSG:32616".
#> ℹ Unit of measurement: metre.
#> n kde geometry
#> Min. : 0.0000 Min. : NA POLYGON :2926
#> 1st Qu.: 0.0000 1st Qu.: NA epsg:4326 : 0
#> Median : 0.0000 Median : NA +proj=long...: 0
#> Mean : 0.7673 Mean :NaN
#> 3rd Qu.: 1.0000 3rd Qu.: NA
#> Max. :28.0000 Max. : NA
#> NA's :2926
# But the KDE estimates are as expected when the data are first transformed to
# a projected CRS
hotspot_dual_kde(
sf::st_transform(memphis_robberies, "EPSG:6410"),
sf::st_transform(memphis_population, "EPSG:6410")
) |>
summary()
#> Cell size set to 500 metres automatically
#> Bandwidth set automatically based on rule of thumb.
#> ℹ Bandwidth for `x` = 5,588 metres.
#> Bandwidth set automatically based on rule of thumb.
#> ℹ Bandwidth for `y` = 5,178 metres.
#> n kde geometry
#> Min. : 0.0000 Min. :0.02357 POLYGON :3300
#> 1st Qu.: 0.0000 1st Qu.:0.14901 epsg:6410 : 0
#> Median : 0.0000 Median :0.23398 +proj=lcc ...: 0
#> Mean : 0.6803 Mean :0.23669
#> 3rd Qu.: 1.0000 3rd Qu.:0.28704
#> Max. :24.0000 Max. :1.10445Created on 2025-08-01 with reprex v2.1.1
It might also be worth checking at this point if hotspot_dual_kde() also needs to check that the two datasets overlap. This is done for x and grid because of #39 but probably also needs to be done for x and y (and possibly for y and grid).