How to make geom_errorbar() work after using coord

I am trying to display error bars using geom_errorbar() after using coord_trans() to plot the x-data in semilog scale. Somehow geom_errorbar() does not work but geom_linerange() does.

My code is as follows:

data <- data.frame(x=c(1, 10, 25, 50, 100, 500, 1000),
                  y=c(0.3990093, 0.6062332, 0.7663158, 0.9779979, 1.6546543, 8.2825181, 18.2218369),
                  std=c(0.09528301, 0.03620494, 0.01246995, 0.06584974, 0.15254007, 0.29668775, 0.86524986))

data %>% 
  ggplot() 
  geom_point(aes(x, y), colour="blue") 
  coord_trans(x = "log10") 
  geom_smooth(aes(x, y), method = "lm", formula = y ~ x, lwd=0.75, se = F, colour="lightblue") 
  #geom_errorbar(aes(x=x, ymin=y-std, ymax=y std), colour="black")
  geom_linerange(aes(x=x, ymin=y-std, ymax=y std), colour="black")

However, geom_linerange() only displays a vertical bar and I would like an horizontal bar on top of the vertical bar, as geom_errorbar() usually allows to display.

Does anyone know how to do that? Thank you in advance!

Estelle

CodePudding user response：

Seems to work ok if we switch the transformation-plotting order by putting the transformation into scale_x_continuous instead of using coord_trans.

data %>% 
  ggplot() 
  geom_point(aes(x, y), colour="blue") 
  scale_x_continuous(trans = scales::log10_trans())  
  geom_smooth(aes(x, y), method = "lm", formula = y ~ x, lwd=0.75, se = F, colour="lightblue") 
  geom_errorbar(aes(x=x, ymin=y-std, ymax=y std), colour="black", width = 0.05) 
  geom_linerange(aes(x=x, ymin=y-std, ymax=y std), colour="black")

CodePudding user response：

A collegue and I found a solution.

It works if a width parameter is specified. As the x-axis is in log-scale, the width parameter needs to be a function of the x-variable, otherwise you won't be able to see it for the values with a higher x-component.

geom_errorbar(aes(x=x, ymin=y-std, ymax=y std), colour="black", width=0.1*data$x)