Add average lines for categorical scatterplots with single group values

Question

I have a categorical scatter plot like this:

which I generated in R with the following code (using the ggplot2 library):

data <- runif(50, 13, 17)
factors <- as.factor(sample(1:3, 50, replace = TRUE))
groups <- as.factor(sample(1:3, 50, replace = TRUE))
data_table <- data.frame(data, factors)
g <- ggplot(data_table, aes(y = data_table[, 1], x = data_table[, 2], colour = groups)) + geom_point(size = 1.5)

I am trying to add an average line for each x-group, but I can't manage to find the right way. I have already tried with the procedure described in this question, but it doesn't work, I reckon because my x-groups are composed of a single x-value each, for which I believe the procedure should be different.

More in detail, if I add:

+ geom_line(stat = "hline", yintercept = "mean", aes(colour = data_table[, 2]))

to the previous code line, it gives me the following error: geom_path: Each group consist of only one observation. Do you need to adjust the group aesthetic?.

If I try with the procedure suggest in the answer to that question, by adding:

+ geom_errorbar(stat = "hline", yintercept = "mean", width=0.8, aes(ymax=..y..,ymin=..y..))

to my initial code (I have removed the geom_jitter(position = position_jitter(width = 0.4)) piece of code, because it added random points to my data plot), I get three lines for each group (each corresponding to the mean of the three groups indicated in red, green, blue for that specifical x-group), as shown in this picture:

Does anyone have any suggestion on how to fix this?

Thank you.

Answer 1

The following code should give you the desired result:

# creating reproducible data
set.seed(1)
data <- runif(50, 13, 17)
factors <- as.factor(sample(1:3, 50, replace = TRUE))
groups <- as.factor(sample(1:3, 50, replace = TRUE))
data_table <- data.frame(data, factors, groups)

# creating the plot
ggplot(data=data_table, aes(x=factor(factors), y=data, color=groups)) + 
  geom_point() +
  geom_errorbar(stat = "hline", yintercept = "mean", width=0.6, aes(ymax=..y.., ymin=..y.., group=factor(factors)), color="black")

which gives:

Checking whether the means are correct:

> by(data_table$data, data_table$factors, mean)
data_table$factors: 1
[1] 15.12186
------------------------------------------------------------------------------------------------- 
data_table$factors: 2
[1] 15.03746
------------------------------------------------------------------------------------------------- 
data_table$factors: 3
[1] 15.24869

which leads to the conclusion that the means are correctly displayed in the plot.

---

Following the suggestion of @rrs, you could also combine it with a boxplot:

ggplot(data=data_table, aes(x=factor(factors), y=data, color=groups)) + 
  geom_boxplot(aes(middle=mean(data), color=NULL)) +
  geom_point(size=2.5)

which gives:

However, the middle line represents the median and not the mean.