[SOLVED] How to fill missing values based on neighboring polygons in R?

How to fill missing values based on neighboring polygons in R?

Suppose the following dataset and add some missing values (only for illustration):

library(dplyr)
library(sf)

demo(nc, ask = FALSE, verbose = FALSE)
nc$AREA[c(30, 45)] <- NA

I can get all the neighboring polygons for each county:

nc %>% mutate(
   INTERSECT = purrr::map(.x = geometry, .f = st_intersects, y = st_geometry(nc))
)

This gives me a list of the indices of neighboring counties for each row. Now I'd like to fill the missing area values with the mean area of neighboring polygons. How would I use these indices to take the mean over the corresponding rows?

Solution

index <- st_touches(nc, nc)

output <- nc %>% 
  mutate(AREA = ifelse(is.na(AREA),
                             apply(index, 1, function(i){mean(.$AREA[i])}),
                             AREA))

output$AREA[c(30, 45)]

[1] 0.1510 0.1335

Checking the answers:

Indices for the two polygons' neighbors.

index[c(30, 45)]

[[1]]
[1] 13 14 29 37 48

[[2]]
[1] 44 87

Find the areas manually.

mean(output$AREA[index[[30]]])

[1] 0.151

mean(output$AREA[index[[45]]])

[1] 0.1335