When should I use geom_map?

Question

I'm making a choropleth map with added points in ggplot. So, following the last example of the geom_map help docs

I came up with this:

require(ggplot2)
require(maps)

set.seed(47)
county_map <- map_data("county", "washington")
names(county_map)[5:6] <- c("state", "id")
countyData <- data.frame(id = unique(county_map$id), value = rnorm(39)) 
map1 <- ggplot(countyData, aes(map_id = id)) +
    geom_map(aes(fill = value), map = county_map, colour = "black") +
    coord_map() +
    expand_limits(x = county_map$long, y = county_map$lat)

print(map1)

which works great for the choropleth map. (Aside that I'm thrilled with the map_data function.) But then I try and add points

pointData <- structure(list(xx = c(-119.872483243387, -122.809038239929, 
-122.143143065312
), yy = c(48.1320425447619, 46.7352071436801, 47.9911548514037
)), .Names = c("xx", "yy"), row.names = c(1746L, 7281L, 2692L
), class = "data.frame")

map1 + geom_point(mapping = aes(xx, yy), data = pointData)

And I can't get it to work. I tried a lot of variations, setting group to NULL, naming aes arguments, etc. No luck. So I find this question which does the exact same thing without a problem by merging the map data with the choropleth data and using geom_polygon, which seems more straightforward anyway. (It took me a little while to work out the ID mapping in the first place because I didn't realize I had to remove the region column name to successfully use id. And the syntax of the first method still seems weird to me.)

So, two questions:

1. How is it possible to add points from a different data frame using the method shown above with geom_map?
2. More importantly, are there any advantages to using geom_map as opposed to the geom_polygon approach?

Answer 1

Your immediate problem is that ggplot has no way to tie your point data to the map. Looking at your data frames, you have this for your map:

str(countyData)
'data.frame':   39 obs. of  2 variables:
 $ id   : chr  "adams" "asotin" "benton" "chelan" ...
 $ value: num  1.995 0.711 0.185 -0.282 0.109 ...

...and this for your points:

str(pointData)
'data.frame':   3 obs. of  2 variables:
 $ xx: num  -120 -123 -122
 $ yy: num  48.1 46.7 48

Do you see any common variables there that would allow ggplot to locate your points?

Still, the problem is easily resolved. I typically use geom_polygon rather than geom_map but that's largely out of habit. This works, for example:

colnames(pointData) <- c('long','lat') # makes consistent with county_map
pointData$group <- 1 # ggplot needs a group to work with
county_map$value <- sapply(1:nrow(county_map),
                           function(x) round(runif(1, 1, 8), 0)) # for colours

ggplot(county_map, aes(x = long, y = lat, group = group)) +
    geom_polygon(aes(fill = value)) +
    coord_map() +
    geom_point(data = pointData, aes(x = long, y = lat), shape = 21, fill = "red")

Which gives the following (note the points).

However, as to whether you should use geom_map or geom_polygon, I have not really thought about the issue much. Maybe somebody else has a view.

Answer 2

This works for me :

map1 <- ggplot(countyData) +
  geom_map( map = county_map, aes(map_id = id,fill = value), 
            colour = "black") + coord_map() +
  expand_limits(x = county_map$long, y = county_map$lat)
  map1 + geom_point(mapping = aes(xx, yy), data = pointData)

For me geom_map is a wrapper of a geom_polygon. It is a layer that contains all geographical settings (lat and long grouped by id).

I would use geom_map when I plot a map and geom_polygon to plot any polygon type.

EDIT To add the map