Visualize your Experimental Results with a dotplot in ggplot2
Published:
How to generate a quick dotplot of your data in R with ggplot2
Here’s a template I use to generate something better than barcharts to visualize differences in response between groups.
First, let’s simulate some data to visualize.
set.seed(42) # set seed for reproducability
df <- data.frame(
group = rep(LETTERS[1:3], each = 6),
value = rnorm(18, mean = 4)
)
Notice the data is in tidy format, if your research was recorded like most lab experiments, it might be in wide format. Thankfully, the dplyr package has the pivot_longer function which makes converting your data from wide to long format a breeze.
Here is the code I use to create a quick dotplot:
library(ggplot2)
my_plot <- ggplot(df, aes(x = group, y = value)) + # these are column names of my dataset
geom_dotplot(
binaxis = "y", # this is probably what you want
stackdir = "center", # for symmetry
dotsize = 0.8, # adjust along with bindwidth to get desired size
fill = "white", # this makes the dots circles instead
binwidth = 0.1 # specifies bin width. Defaults to 1/30 of the range of the data
) +
stat_summary(
fun.y = mean, # gives horizontal line to compare, can easily change to median
geom = "errorbar", # make the line appear
aes(
ymax = after_stat(y), # upper bound of line
ymin = after_stat(y), # lower bound of line
x = group # one line per group
),
width = 0.5, # how wide the line should be
linetype = 1 # solid or 2 for dashed.
) +
scale_x_discrete(labels = c("Control", "Alfalfa", "Wheat")) + # if x-axis groups need renamed
labs(title = "Example Dotplot Template", # give your desired title
x = "Treatment", # label the x axis
y = "Value") + # label the y axis
theme_classic() # a ggplot theme
my_plot
And there we have it! This can be modified to get what you need, but I hope it is helpful for scientists wanting to visualize their data.
That’s a great baseline for getting a feel for how the data looks, but you’ll likely want to test for significance for any differences that you might see.
Testing for significance
Fortunately, we can within the ggplot ecosystem test for differences. The code I use to do so is below. I start with the my_plot we made in the code above.
library(ggpubr)
my_plot +
ggpubr::stat_compare_means(comparisons =
list(c("A", "B"),
c("B", "C"),
c("A", "C")),
method = "t.test",
label = "p.signif") + # "p.format" if you want the actual p-value.
ggpubr::stat_compare_means(label.y = min(df$value), method = "anova") # add ANOVA p-value.
And there we have it!