CODEWITHSUNDEEP
rggplot2
Dekike
Dekike

Reputation: 1284

How to scale the errorbars to the dimensions of the right y-axis in a plot with one x-axis and two y-axis

I have a dataframe df1 as below.

df1<-data.frame(Hour=c(0,3,6,9,12,15,18,21),
                n=c(18426,19345,20123,21450,23456,23510,21453,18456),
                mean=c(23.9234, 34.9456,28.9891,44.6452,47.4567,38.9483,34.9632,29.8765),
                ci=c(4.2345,6.3345,12.1345,17.3445,13.1545,12.1745,10.1945,28.2445))
df1$Hour<-as.factor(df1$Hour)
df1

  Hour     n    mean      ci
1    0 18426 23.9234  4.2345
2    3 19345 34.9456  6.3345
3    6 20123 28.9891 12.1345
4    9 21450 44.6452 17.3445
5   12 23456 47.4567 13.1545
6   15 23510 38.9483 12.1745
7   18 21453 34.9632 10.1945
8   21 18456 29.8765 28.2445

I created the plot that I show below after some search on the internet, with one x-axis and two y-axis. The left y-axis showing the variable n and the right y-axis showing the variable mean. The "orange" line represents the confidence interval (ci) for the mean (right y-axis).

ggplot() + 
  geom_bar(mapping = aes(x = df1$Hour, y = df1$n), stat = "identity", fill = "grey") +
  geom_line(mapping = aes(x = df1$Hour, y = df1$mean*494.2611, group=1 ), size = 1, color = "blue") + # 525.3868 result from the division of 23456/44.6452
  geom_point(mapping = aes(x = df1$Hour, y = df1$mean*494.2611 )) + 
  scale_y_continuous(name = "n", 
                     sec.axis = sec_axis(~./494.2611, name = "Mean")) + 
  theme(
    axis.title.y = element_text(color = "darkgrey"),
    axis.title.y.right = element_text(color = "blue")) + 
  labs(x = "Hour") +
  geom_errorbar(mapping= aes(x = df1$Hour, ymin = df1$mean*494.2611 - df1$ci, ymax = df1$mean*494.2611 + df1$ci),
                position = position_dodge(0.9),
                width = 0.4, colour = "orange", 
                alpha = 0.9, size = 0.5)

enter image description here

The problem is that the errorbars appear too small compare to the scale of the variable mean, and hence all look the same way. I think that the errorbars are scaled to the left y-axis.

Does anyone know where is the mistake?

Upvotes: 0

Views: 438

Answers (1)

January
January

Reputation: 17110

You forgot your parentheses. Instead of 

 ymin = df1$mean*494.2611 - df1$ci

use 

 ymin = (df1$mean - df1$ci) * 494.2611

(same for ymax)

Explanation: dual scales in ggplot is a hack. You have to manually rescale the data to fit the primary scale. For error bars, you need ymin and ymax, and these are the values that you need to scale. But ymin=df1$mean - df1$ci, so you need to multiply the whole value, not just the mean.

Upvotes: 1

Related Questions