CODEWITHSUNDEEP
rggplot2
llewmills
llewmills

Reputation: 3568

Suppressing some crossbars when drawing error bars with ggplot2

I would like to add an if else statement within the mapping statement within geom_errorbar() that removes the lower error bar if it fulfills a certain criteria. See the following toy dataset and graph

df <- data.frame(change = rep(c("increased", "reduced", "same"), 2),
                 group = rep(c("0", "1"), each = 3),
                 freq = c(0, 17, 21, 1, 27, 9),
                 perc = c(0, 44.73, 55.26, 2.70, 72.97, 24.32),
                 se = c(NaN, 12.06, 10.85, 16.22, 8.54, 14.30))

polytPlot <- ggplot(dfT, aes(y = perc, x = change, fill = Group)) +
                    geom_bar(stat = "identity", position = "dodge") +
                    scale_y_continuous(breaks=pretty) + 
                    ylab("% of group total") +
                    xlab("Change") +
                    geom_errorbar(aes(ymin = ifelse(perc-se < 0, 0, perc-se), ymax = perc+se), position = position_dodge(.9), width = .1)
polytPlot

Note that the ifelse() statement in the ymin argument of geom_errorbar in the above example does work, reducing the lower error bar to zero, but it still displays the horizontal portion of the error bar. How do I suppress this, so only the upper error-bar appears? I tried entering NULL instead of 0 in my conditional ifelse() statement but got an error message. Perhaps a conditional argument to width =?

Upvotes: 4

Views: 1213

Answers (1)

Claus Wilke
Claus Wilke

Reputation: 17810

One option is to remove the horizontal crossbars altogether:

ggplot(df, aes(y = perc, x = change, fill = group)) +
  geom_bar(stat = "summary", position = "dodge") +
  scale_y_continuous(breaks=pretty) + 
  ylab("% of group total") +
  xlab("Change") +
  geom_linerange(aes(ymin = ifelse(perc-se < 0, 0, perc-se), ymax = perc+se), position = position_dodge(.9))

enter image description here

If you really want the crossbars, however, you can draw them separately using geom_segment():

ggplot(df, aes(y = perc, x = change, fill = group)) +
  geom_bar(stat = "summary", position = "dodge") +
  scale_y_continuous(breaks=pretty, limits = c(0, 85)) + 
  ylab("% of group total") +
  xlab("Change") +
  geom_linerange(aes(ymin = ifelse(perc-se < 0, 0, perc-se), ymax = perc+se), position = position_dodge(.9)) +
  geom_segment(aes(x = as.numeric(change) + .45*(as.numeric(group)-1.5) - .05,
                   xend = as.numeric(change) + .45*(as.numeric(group)-1.5) + .05,
                   y = perc + se, yend = perc + se)) +
  geom_segment(aes(x = as.numeric(change) + .45*(as.numeric(group)-1.5) - .05,
                   xend = as.numeric(change) + .45*(as.numeric(group)-1.5) + .05,
                   y = perc - se, yend = perc - se))

enter image description here

Note that the missing segment is removed by the limits statement I added to scale_y_continuous().

Upvotes: 5

Related Questions