Overlaying boxplot with histogram in ggplot2

Question

Hi I want to create a similar chart as shown below with R script:

taken from: https://community.tableau.com/thread/194440

this is my code in R :

library(ggplot2)

ifile <- read.table("C:/ifiles/test.txt", skip = 2, header = TRUE, sep="\t")
ifileVI <- data.frame(ifile["VI"], ifile["Site"])
x<-quantile(ifileVI$VI,c(0.01,0.99))
data_clean <- ifileVI[bfileVI$VI >=x[1] & ifileVI$VI <=x[2],]

p <- ggplot(data_clean, aes(x = Site, y = VI, group=Site)) + geom_boxplot() + geom_histogram(binwidth = 0.05)

p

however im getting the following error:

Error: stat_bin() must not be used with a y aesthetic.

bfileVI:

Id	    VI	Site
WFR1	2.91	1
WFR1	2.89	2
WFR1	2.86	3
WFR1	2.91	4
WFR1	2.87	1
WFR1	2.67	2
WFR1	2.76	3
WFR1	2.74	4
WFR1	2.98	4
WFR1	2.89	3
WFR1	2.55	4
WFR1	2.96	3
WFR1	2.71	1
WFR1	2.98	2
WFR1	2.89	3
WFR2	2.55	2
WFR2	2.86	4
WFR2	2.91	3
WFR2	287	1
WFR2	2.74	2
WFR2	2.98	1
WFR2	2.89	2
WFR2	2.55	3
WFR2	2.96	4
WFR2	2.71	1
WFR2	2.86	2
WFR2	2.91	3
WFR2	287	4
WFR2	2.67	1
WFR2	2.76	2
WFR2	2.74	3
WFR2	2.98	4
WFR2	2.89	1
WFR2	2.55	2
WFR2	2.96	3
WFR2	2.71	4
WFR2	2.98	1
WFR2	2.89	2
WFR2	2.55	3
WFR2	2.86	4

Answer 1

You're getting Error: stat_bin() must not be used with a y aesthetic. because you can't specify y in the aesthetic of a histogram. If you want to mix plots that have different parameters, you need to supply distinct aesthetics. I'll demonstrate with iris like so:

ggplot(iris, aes(x = Sepal.Width)) + 
  geom_histogram(binwidth = 0.05) +
  geom_boxplot(aes(x = 3, y = Sepal.Width))

Unfortunately, the default for boxplots is vertical, for histograms is horizontal, and coord_flip() is all-or-nothing, so you're left with this awful thing:

Best I can figure out is instead of having them overlap, put one on top of the other with the gridExtra package:

a <- ggplot(iris, aes(x = Sepal.Width)) + 
  geom_histogram(binwidth = 0.05) 

b <- ggplot(iris, aes(x = "", y = Sepal.Width)) + 
  geom_boxplot() + 
  coord_flip()

grid.arrange(a,b,nrow=2)

which gives us something pretty good:

Answer 2

You can try to replace histogram with rectangles to generate a plot like this:

---

How to do this:

Generate random data

df <- data.frame(State = LETTERS[1:3],
                 Y = sample(1:10, 30, replace = TRUE),
                 X = rep(1:10, 3))

Replace histogram with rectangles

library(ggplot2)

# You can plot geom_histogram or bar (pre-counted stats)
ggplot(df, aes(X, Y)) +
    geom_bar(stat = "identity", position = "dodge") +
    facet_grid(State ~ .)
# Or you can plot similar figure with geom_rect
ggplot(df)  +
    geom_rect(aes(xmin = X - 0.4, xmax = X + 0.4, ymin = 0, ymax = Y)) +
    facet_grid(State ~ .)

Add boxplot

To add boxplot we need to:

1. Flip coordinates (function coord_flip)
2. Switch X and Y values in geom_rect

Code:

ggplot(df)  +
    geom_rect(aes(xmin = 0, xmax = Y, ymin = X - 0.4, ymax = X + 0.4)) +
    geom_boxplot(aes(X, Y)) +
    coord_flip() +
    facet_grid(State ~ .)

Result:

Final plot code with nicer visuals

ggplot(df)  +
    geom_rect(aes(xmin = 0, xmax = Y, ymin = X - 0.4, ymax = X + 0.4),
              fill = "blue", color = "black") +
    geom_boxplot(aes(X, Y), alpha = 0.7, fill = "salmon2") +
    coord_flip() +
    facet_grid(State ~ .) +
    theme_classic() +
    scale_y_continuous(breaks = 1:max(df$X))

Answer 3

A good example where ggstance package comes in handy.

library(ggplot2)
library(ggstance)

ggplot(iris, aes(x = Sepal.Length)) +
  geom_histogram() +
  geom_boxploth(aes(y = 3), width = 2, color = "blue", lwd = 2, alpha = .5) +
  theme_minimal() +
  facet_wrap(~Species, ncol = 1)