Question

The graph I’m currently trying to make falls a little between two stools. I want to make a histogram that is composed of stacked and labelled boxes. Here’s an example of exactly the sort of thing I’m talking about, taken from a recent article in the New York Times:

http://farm8.staticflickr.com/7109/7026409819_1d2aaacd0a.jpg

Is it possible to achieve this using ggplot2?

To amplify the question somewhat, so far what I have is:

dfr <- data.frame(
name = LETTERS[1:26],
percent = rnorm(26, mean=15)
)

ggplot(dfr, aes(x=percent, fill=name)) + geom_bar() +
stat_bin(geom="text", aes(label=name))

…which I’m clearly doing all wrong. Ultimately what I’d ideally like is something along the lines of the manually-modified graph below, with (say) letters A to M filled one shade and N to Z filled another.

http://farm8.staticflickr.com/7116/7026536711_4df9a1aa12.jpg

Answer 1

Here you go!

set.seed(3421)
# added type to mimick which candidate is supported
dfr <- data.frame(
  name    = LETTERS[1:26],
  percent = rnorm(26, mean=15),
  type    = sample(c("A", "B"), 26, replace = TRUE)
)

# easier to prepare data in advance. uses two ideas
# 1. calculate histogram bins (quite flexible)
# 2. calculate frequencies and label positions
dfr <- transform(dfr, perc_bin = cut(percent, 5))
dfr <- ddply(dfr, .(perc_bin), mutate, 
  freq = length(name), pos = cumsum(freq) - 0.5*freq)

# start plotting. key steps are
# 1. plot bars, filled by type and grouped by name
# 2. plot labels using name at position pos
# 3. get rid of grid, border, background, y axis text and lables
ggplot(dfr, aes(x = perc_bin)) +
  geom_bar(aes(y = freq, group = name, fill = type), colour = 'gray', 
    show_guide = F) +
  geom_text(aes(y = pos, label = name), colour = 'white') +
    scale_fill_manual(values = c('red', 'orange')) +
    theme_bw() + xlab("") + ylab("") +
    opts(panel.grid.major = theme_blank(), panel.grid.minor = theme_blank(),
      axis.ticks = theme_blank(), panel.border = theme_blank(), 
      axis.text.y = theme_blank())