[R] How to overlay points over boxplots

A pure boxplot

First, let’s make a boxplot. We will use the data set ToothGrowth coming with R.

dat<-ToothGrowth
dat$dose<-as.factor(dat$dose) # convert the dose to a factor
plt<-ggplot(dat, aes(x=dose, y=len, fill=supp)) + theme_bw()
plt<-plt + geom_boxplot()
plt

add points with geom_jitter()

Now let’s add points over the boxplots with the function geom_jitter(). Here we use the function position_jitter() to control the points’ spreads.

plt + geom_jitter(position = position_jitter(width=0.4), pch=24, size=2, alpha=0.6)

This doesn’t look right, because the points are superimposed over both boxes in each dose level; they are supposed to be separated by the factor ‘supp’.

Now let’s try a different function position_jitterdodge()

plt + geom_jitter(position = position_jitterdodge(jitter.width=0.4), pch=24, size=2, alpha=0.6)

This looks good, right. Note that we used ‘fill’ to control the colors of both boxplots and points, so the selected points by ‘pch’ should have a property of fill.

Add points with geom_point()

We can also use geom_point() plus position_jitterdodge() to make similar plots, as below:

plt + geom_point(position = position_jitterdodge(jitter.width=0.4), pch=24, size=2, alpha=0.6)

The function position_jitterdodge() provides jitter effects, and the degree of spread can be tuned with the parameter jitter.width, default is 0.4. Let’s try a smaller number 0.2.

plt + geom_point(position = position_jitterdodge(jitter.width=0.2), pch=24, size=2, alpha=0.6)

As you can see, the points are clustered more narrowly in the horizontal direction.

The categorical variable should be a factor

There is a caveat here. The variable used for specifying filling color must be a factor, otherwise the points will not be grouped correctly.

Let’s create a new variable supp2, which is a character version of supp, and use supp2 for color fill.

dat$supp2<-as.character(dat$supp)
plt<-ggplot(dat, aes(x=dose, y=len, fill=supp2)) + theme_bw()
plt<-plt + geom_boxplot()
plt + geom_point(position = position_jitterdodge(jitter.width=0.4), pch=24, size=2, alpha=0.6)

As you can see, the points are spreaded out, even though the same parameter jitter.width was given. Let’s try a smaller value 0.2

plt + geom_point(position = position_jitterdodge(jitter.width=0.2), pch=24, size=2, alpha=0.6)

This helps a bit.

Conclusions

To make boxplot with superimposed points, one need use the function position_jitterdodge() to specify point positions, with this both geom_point() and geom_jitter() can be used.

Also the group variables used for specifying colors or filling colors should be a factor, not character; otherwise some unexpected behavior may happen.

Happy Programming 😄 👍

References

• ggplot2 boxplot quick start: http://www.sthda.com/english/wiki/ggplot2-box-plot-quick-start-guide-r-software-and-data-visualization