0

In a ggplot graph I want to set limits to the y on the first axis but I have problem with the secondary axis vizualisation.

Here is what the code to generate the graph (inspired from this post):

mydata <- data.frame(group = c("A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K"), 
                    count = c(140, 90, 25, 10, 10, 9, 8, 7, 7, 7, 6)) %>%
  mutate(cumsum=cumsum(count)) %>%
  mutate(cumperc=cumsum / sum(count) *100)
mydata$group <- factor(mydata$group, levels = mydata$group)

scaleRight <- tail(mydata$cumperc, n=1)/head(mydata$count, n=1)

ggplot(mydata, aes(x=group)) +
  geom_bar(aes(y=count), fill='deepskyblue4', stat="identity") +
  geom_path(aes(y=cumperc/scaleRight, group=1),colour="orange", size=0.9) +
  geom_point(aes(y=cumperc/scaleRight, group=1),colour="orange") +
  scale_y_continuous(expand = c(0, 0), sec.axis = sec_axis(~.*scaleRight, name = "Cumulative percent (%)", 
                                                           breaks=seq(0,100,10))) +
  coord_cartesian(ylim = c(0, 650)) +
  theme_classic() +
  theme(axis.text.x = element_text(angle=90, vjust=0.6),
        axis.text=element_text(size=12),
        axis.title=element_text(size=14,face="bold")) +
  labs( y="Number of person", x="group")

My problem is that the second axis is squished.

plot output

Any suggestion ?

Another thing, I would like to add an horizontal line base on the secondary axis (at 90%). I tried with geom_hline but it take the first axis into account.

Here is the graph I want to obtain (done in excel) enter image description here

PIFASTE
  • 87
  • 2
  • 8
  • why have you manually set y limits with `coord_cartesian(ylim = c(0, 650))` ? Surely removing that would give your desired result? – Chris Apr 30 '20 at 10:28
  • Yes but I have another plot and I wanted to have the same limits between both for comparison purpose – PIFASTE Apr 30 '20 at 10:54
  • So what's your expected output? Do you want the line plot to reach up to the top of the axis? – Chris Apr 30 '20 at 10:56
  • I have just edited my post to add expected results – PIFASTE Apr 30 '20 at 12:32

1 Answers1

1

You just needed to adjust the scale adjustment in scaleRight to use the full y axis extent (up to 700 in your example) instead of just the range of data for the bars.

I've added the variable ymax to make it easy to change the range of data we want to extend our axis over.

The only other adjustment I made was to include expand = expansion(mult = c(0,.05)) to replace expand=c(0,0) to add a 5% padding to the top of the axis as previously the top of the line was being cropped. 

library(tidyverse)
mydata <- data.frame(group = c("A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K"), 
                     count = c(140, 90, 25, 10, 10, 9, 8, 7, 7, 7, 6)) %>%
  mutate(cumsum=cumsum(count)) %>%
  mutate(cumperc=cumsum / sum(count) *100)
mydata$group <- factor(mydata$group, levels = mydata$group)

ymax <- 700
scaleRight <- max(mydata$cumperc)/ymax

ggplot(mydata, aes(x=group)) +
  geom_bar(aes(y=count), fill='deepskyblue4', stat="identity") +
  geom_path(aes(y=cumperc/scaleRight, group=1),colour="orange", size=0.9) +
  geom_point(aes(y=cumperc/scaleRight, group=1),colour="orange") +
  geom_hline(aes(yintercept=90/scaleRight), linetype='dashed')+
  scale_y_continuous(expand = expansion(mult = c(0,.05)),
                     sec.axis = sec_axis(~.*scaleRight, name = "Cumulative percent (%)",
                                         breaks=seq(0,100,10))) +
  coord_cartesian(ylim = c(0, ymax)) +
  theme_classic() +
  theme(axis.text.x = element_text(angle=90, vjust=0.6),
        axis.text=element_text(size=12),
        axis.title=element_text(size=14,face="bold")) +
  labs( y="Number of person", x="group")

Created on 2020-04-30 by the reprex package (v0.3.0)

Chris
  • 3,836
  • 1
  • 16
  • 34