Density2d Plot using another variable for the fill (similar to geom_tile)?

Question

I am trying to plot a map for my final project, and I am trying to do a heat map of crime by BLock in the US.

For each block, I have Lat, Lon, and a prediction of th

Answer 1

First, let's load the data:

data<-read.csv(file = "NY subsample.csv")

Data points

Then, let's try just plotting the basic locations and values of the data:

require('ggplot2')
# start with points
pred.points <- ggplot(data = data,
       aes(x = GEO_CENTROID_LON,
           y = GEO_CENTROID_LAT,
           colour = prediction)) + 
  geom_point()
print(pred.points)
ggsave(filename = "NYSubsamplePredPoints.png",
       plot = p2,
       scale = 1,
       width = 5, height = 3,
       dpi = 300)

which gives us this:

Binned data

Then, you can try to plot the mean in a 2-D region using stat_summary2d():

pred.stat <- ggplot(data = data,
                      aes(x = GEO_CENTROID_LON,
                          y = GEO_CENTROID_LAT,
                          z = prediction)) + 
  stat_summary2d(fun = mean)
print(pred.stat)
ggsave(filename = "NYSubsamplePredStat.png",
       plot = pred.stat,
       scale = 1,
       width = 5, height = 3,
       dpi = 300)

which gives us this plot of the mean value of prediction in each box.

Binned and with custom colormap and correct projection

Next, we can set the bin size, color scales, and fix the projection:

# refine breaks and palette ----
require('RColorBrewer')
YlOrBr <- c("#FFFFD4", "#FED98E", "#FE9929", "#D95F0E", "#993404")
pred.stat.bin.width <- ggplot(data = data,
                    aes(x = GEO_CENTROID_LON,
                        y = GEO_CENTROID_LAT,
                        z = prediction)) + 
  stat_summary2d(fun = median, binwidth = c(.05, .05)) + 
  scale_fill_gradientn(name = "Median",
                       colours = YlOrBr,
                       space = "Lab") +
  coord_map()
print(pred.stat.bin.width)
ggsave(filename = "NYSubsamplePredStatBinWidth.png",
       plot = pred.stat.bin.width,
       scale = 1,
       width = 5, height = 3,
       dpi = 300)

which gives us this:

Plotted over a base map

And last of all, here's the data overlain on a map.

require('ggmap')
map.in <- get_map(location = c(min(data$GEO_CENTROID_LON),
                               min(data$GEO_CENTROID_LAT),
                               max(data$GEO_CENTROID_LON),
                               max(data$GEO_CENTROID_LAT)),
                  source = "osm")
theme_set(theme_bw(base_size = 8))
pred.stat.map <- ggmap(map.in) %+% data + 
  aes(x = GEO_CENTROID_LON,
      y = GEO_CENTROID_LAT,
      z = prediction) +
  stat_summary2d(fun = median, 
                 binwidth = c(.05, .05),
                 alpha = 0.5) + 
  scale_fill_gradientn(name = "Median",
                       colours = YlOrBr,
                       space = "Lab") + 
  labs(x = "Longitude",
       y = "Latitude") +
  coord_map()
print(pred.stat.map)
ggsave(filename = "NYSubsamplePredStatMap.png",
       plot = pred.stat.map,
       scale = 1,
       width = 5, height = 3,
       dpi = 300)

Setting the colormap

And finally, to set the colormap to something like http://www.cadmaps.com/images/HeatMapImage.jpg, we can take a guess at the colormap:

colormap <- c("Violet","Blue","Green","Yellow","Red","White")

and do the plotting again:

pred.stat.map.final <- ggmap(map.in) %+% data + 
  aes(x = GEO_CENTROID_LON,
      y = GEO_CENTROID_LAT,
      z = prediction) +
  stat_summary2d(fun = median, 
                 binwidth = c(.05, .05),
                 alpha = 1.0) + 
  scale_fill_gradientn(name = "Median",
                       colours = colormap,
                       space = "Lab") + 
  labs(x = "Longitude",
       y = "Latitude") +
  coord_map()
print(pred.stat.map.final)