Width of error bars in line plot using ggplot2

Question

I have some data with standard errors associated and would like to display these with error bars. That's what I have:

# generate some data
hod <- data.frame(h = c(1:24,1:24,1:24), mean = 1:(24*3) + runif(24*3, 0, 5),ci = runif(24*3, 0, 2), t = c(rep("a",24),rep("b",24),rep("c",24)))

pd <- position_dodge(0.3)
  dayplot <- ggplot(hod, aes(x=h, y=mean, colour=as.factor(t),group=as.factor(t))) + 
    geom_line(position=pd, size=1) +
    geom_errorbar(aes(ymin=mean-ci, ymax=mean+ci),
                  width=1,
                  size=0.5,
                  position=pd) +
    geom_point(position=pd, shape=21, size=1, fill="white") +
    scale_x_continuous(limits=c(-0.5,23.5),
                       breaks=c(0:8*3),
                       labels=ifelse(
                              c(0:8*3) < 10,
                              paste('0',c(0:8*3),':00',sep=''),
                              paste(c(0:8*3),':00',sep='')
                              )
                       ) +
    xlab("Hour of day") + ylab(ylabel) + labs(title = varlabels[var]) +
    theme_minimal() + 
    theme(plot.margin = unit(c(1,0,1,1), "cm"), 
          axis.title.x = element_text(vjust=-1),
          axis.title.y = element_text(angle=90, vjust=0),
          legend.margin = unit(c(0), "cm"),
          legend.key.height = unit(c(0.9), "cm"),
          panel.grid.major = element_line(colour=rgb(0.87,0.87,0.87)),
          panel.grid.minor = element_blank(),
          plot.background = element_rect(fill = rgb(0.97,0.97,0.97), linetype=0)
    )

The only thing of interest is probably:

geom_errorbar(aes(ymin=mean-ci, ymax=mean+ci),
                      width=1,
                      size=0.5,
                      position=pd)

It gives: all

Now when I group the data by a factor variable (as.factor(t)), I get several lines instead of one, which is what I want, BUT, as you can see, the horizontal lines at the error bars are more narrow, and I can't figure out why. I tried changing and even taking away the width and size attribute of geom_errorbar, but nothing happens. Is there a way to have the same width of the horizontal lines for every chart, no matter the data? I mean, why should it vary? Or does that width convey some information?

enter image description here

I think it would already help if somebody could explain to me the use and function of width and size for geom_errorbar, as it should be data- and thus problem-independent. — grssnbchr
– grssnbchr, Commented Oct 17, 2013 at 9:11
I tried to reproduce the problem with a simplified example and couldn't. It's not obvious, what is going wrong. If you provide data it's much easier to find the problem. — Roland
– Roland, Commented Oct 17, 2013 at 9:57

thc · Accepted Answer · 2015-03-15 10:30:35Z

18

Below is a reproducible example using random data. The fix to the problem is to multiply the width by the number of classes/factors that you have. In the plot below, since I used three factors, using a width of 3 fixes the problem. ggplot2 seems to calculate the relative width by the number of data points in your dataset, rather than the numeric values on the x-axis. This is (IMO) a bug.

library(ggplot2)
library(grid)

#plot with factors
hod <- data.frame(h = c(1:24,1:24,1:24), mean = 1:(24*3) + runif(24*3, 0, 5),ci = runif(24*3, 0, 2), t = c(rep("a",24),rep("b",24),rep("c",24)))
pd <- position_dodge(0.3)
  dayplot <- ggplot(hod, aes(x=h, y=mean, colour=as.factor(t),group=as.factor(t))) + 

    geom_line(position=pd, size=1) +
    geom_errorbar(aes(ymin=mean-ci, ymax=mean+ci),
                  width=1,
                  size=0.5,
                  position=pd) +
    geom_point(position=pd, shape=21, size=1, fill="white") +
    scale_x_continuous(limits=c(-0.5,23.5),
                       breaks=c(0:8*3),
                       labels=ifelse(
                              c(0:8*3) < 10,
                              paste('0',c(0:8*3),':00',sep=''),
                              paste(c(0:8*3),':00',sep='')
                              )
                       ) +
    xlab("Hour of day") +
    theme_minimal() + 
    theme(plot.margin = unit(c(1,0,1,1), "cm"), 
          axis.title.x = element_text(vjust=-1),
          axis.title.y = element_text(angle=90, vjust=0),
          legend.margin = unit(c(0), "cm"),
          legend.key.height = unit(c(0.9), "cm"),
          panel.grid.major = element_line(colour=rgb(0.87,0.87,0.87)),
          panel.grid.minor = element_blank(),
          plot.background = element_rect(fill = rgb(0.97,0.97,0.97), linetype=0)
    )
print(dayplot)


#plot without factors
hod <- data.frame(h = c(1:24,1:24,1:24), mean = 1:(24) + runif(24, 0, 5),ci = runif(24, 0, 2))
pd <- position_dodge(0.3)
  dayplot <- ggplot(hod, aes(x=h, y=mean)) + 

    geom_line(position=pd, size=1) +
    geom_errorbar(aes(ymin=mean-ci, ymax=mean+ci),
                  width=1,
                  size=0.5,
                  position=pd) +
    geom_point(position=pd, shape=21, size=1, fill="white") +
    scale_x_continuous(limits=c(-0.5,23.5),
                       breaks=c(0:8*3),
                       labels=ifelse(
                              c(0:8*3) < 10,
                              paste('0',c(0:8*3),':00',sep=''),
                              paste(c(0:8*3),':00',sep='')
                              )
                       ) +
    xlab("Hour of day") +
    theme_minimal() + 
    theme(plot.margin = unit(c(1,0,1,1), "cm"), 
          axis.title.x = element_text(vjust=-1),
          axis.title.y = element_text(angle=90, vjust=0),
          legend.margin = unit(c(0), "cm"),
          legend.key.height = unit(c(0.9), "cm"),
          panel.grid.major = element_line(colour=rgb(0.87,0.87,0.87)),
          panel.grid.minor = element_blank(),
          plot.background = element_rect(fill = rgb(0.97,0.97,0.97), linetype=0)
    )

print(dayplot)

answered Mar 15, 2015 at 10:30

thc

9,7151 gold badge29 silver badges44 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

grssnbchr Over a year ago

Now, where lies the difference to my example? You're still using width = 1, as far as I can see. Do you mean you should use width = 3?

thc Over a year ago

Yes. Try your own example with width = 9 (or however many levels you have).

grssnbchr Over a year ago

Thanks. I posted an issue on GitHub github.com/hadley/ggplot2/issues/1068

curious lab rat Over a year ago

This issue seems to remain unresolved for many of us for whom the standard in the field is NOT to scale errorbar width by number of elements. I know that ggplot developers think this is a feature not a bug, but our colleagues/journals beg to differ. @hadley, help? thanks!

thc Over a year ago

@curiouslabrat It's definitely a bug. Check out the github issue by grssnbchr and maybe re-open with a simple rep. ex.

|

cfnerd · Accepted Answer · 2018-12-20 15:42:32Z

5

I have managed to solve a similar issue. In my case I wanted to set both horizontal and vertical errorbar heads to the same size - regardless of the aspect ratio of the plot.

Based on the original posted code:

f <- ggplot_build(dayplot)

f$plot$layers[[5]]$geom_params$width  <- 0.02 * diff(f$layout$panel_params[[1]]$x.range)
f$plot$layers[[6]]$geom_params$height <- 0.02 * diff(f$layout$panel_params[[1]]$y.range)

dayplot <- f$plot

This will set the errorbar head to 2% of the axis range. Maybe could solve your issue.

edited Dec 20, 2018 at 15:42

cfnerd

3,79912 gold badges36 silver badges45 bronze badges

answered Dec 20, 2018 at 15:22

ThiagoMAF

611 silver badge2 bronze badges

Collectives™ on Stack Overflow

Width of error bars in line plot using ggplot2

2 Answers 2

7 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

7 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related