Add legend to “geom_bar” using the ggplot2 package

Question

I am a newbie to R, so please pardon my ignorance. I made a pseudo-stacked barplot in which I drew 4 sets of bars on top of each other using geom_bar. There are 4 health status categories (alive, dead, infected, & sod-dead) for three species of oak trees (QUAG, QUKE, QUCH).

My code is as follows:

---

x <- as.data.frame(list(variable=c("QUAG", "QUKE", "QUCH"), alive = c(627,208,109),  infected = c(102,27,0), dead = c(133,112,12), sod.dead=c(49,8,0)))

x.plot = ggplot(x, aes(variable, alive)) + geom_bar(fill="gray85") + 
  geom_bar(aes(variable,dead), fill="gray65") +
  geom_bar(aes(variable, infected), fill="gray38") +
  geom_bar(aes(variable, sod.dead), fill="black")+
  opts(panel.background = theme_rect(fill='gray100'))
x.plot

---

Now I want to make a legend that shows which shade of gray relates to tree status, ie, "gray65" is "dead trees", etc. I've been trying for the past hour and can't get it to work.

Answer 1

I see that @Brandon Bertelsen has posted a great answer. I would like to add some code that addresses additional details mentioned in the original post:

1. After you reshape your data and map health status to fill , ggplot will create the legend automatically.
2. I suggest using scale_fill_manual() to get the exact grays mentioned in the original post.
3. theme_bw() is a handy function to quickly get a black and white look to your plot.
4. The plotting order of factor levels/colors can be controlled by specifying the desired order with the levels argument of factor() .
5. A dodged barplot (instead of stacked) may have some advantages for this data set.

---

library(reshape2)
library(ggplot2)

x <- as.data.frame(list(variable=c("QUAG", "QUKE", "QUCH"), 
                        alive=c(627, 208, 109),  infected=c(102, 27, 0), 
                        dead=c(133, 112, 12), sod.dead=c(49, 8, 0)))

# Put data into 'long form' with melt from the reshape2 package.
dat = melt(x, id.var="variable", variable.name="status")

head(dat)
#    variable   status value
# 1      QUAG    alive   627
# 2      QUKE    alive   208
# 3      QUCH    alive   109
# 4      QUAG infected   102
# 5      QUKE infected    27
# 6      QUCH infected     0

# By manually specifying the levels in the factor, you can control
# the stacking order of the associated fill colors.
dat$status = factor(as.character(dat$status), 
                    levels=c("sod.dead", "dead", "infected", "alive"))

# Create a named character vector that relates factor levels to colors.
grays = c(alive="gray85", dead="gray65", infected="gray38", sod.dead="black")

plot_1 = ggplot(dat, aes(x=variable, y=value, fill=status)) +
         theme_bw() +
         geom_bar(position="stack") +
         scale_fill_manual(values=grays)

ggsave(plot=plot_1, filename="plot_1.png", height=5, width=5)

# You may also want to try a dodged barplot.
plot_2 = ggplot(dat, aes(x=variable, y=value, fill=status)) +
         theme_bw() +
         geom_bar(position="dodge") +
         scale_fill_manual(values=grays)

ggsave(plot=plot_2, filename="plot_2.png", height=4, width=5)

Answer 2

You need to reshape your data.

library(reshape)
library(ggplot2)

x <- as.data.frame(list(variable=c("QUAG", "QUKE", "QUCH"), alive = c(627,208,109),  infected = c(102,27,0), dead = c(133,112,12), sod.dead=c(49,8,0)))

x <- melt(x)
colnames(x) <- c("Type","Status","value")

ggplot(x, aes(Type, value, fill=Status)) + geom_bar(position="stack")