Add quantiles to geom_density plot with stat_summary

Question

I previously asked a question about using stat_summary to add a vertical line for the median value in my density plot. I'm now trying to adapt this to also add vertical lines for the 95% confidence interval. I've been using another question as the base for my code, but it doesn't add any lines to my plot. How can I get this to work?

library(ggplot2)

set.seed(42)

dat <- data.frame(
  site = rep(LETTERS[1:4], each = 25),
  sex = rep(c("f", "m"), 50),
  mass = rnorm(100, 50, 5)
)

dat |>
  ggplot(aes(x = mass, col = sex)) +
  geom_density() +
  stat_summary(
    aes(xintercept = after_stat(x), y = 0),
    fun = mean, geom = "vline", orientation = "y"
  ) +
  stat_summary(
    aes(xintercept = after_stat(x), y = 0),
    fun.min = function(pred) { quantile(pred, probs = 0.025) }, 
    fun.max = function(pred) { quantile(pred, probs = 0.975) },
    geom = "vline", orientation = "y", linetype = "dashed"
) +
facet_wrap(~site)

Remove orientation = "y". But you'll also have to make other changes to get what you want... [It would have been helpful to report the warning you get with your current code: "Removed 8 rows containing missing values or values outside the scale range (geom_vline())" is a big clue to what's going on.] — Limey
– Limey, Commented Mar 22, 2024 at 18:20
Note the 95% quantiles are not the same thing as a 95% confidence interval (presumably for the mean). How exactly do you want to interpret these lines? — MrFlick
– MrFlick, Commented Mar 22, 2024 at 18:25
@Limey any idea what the other changes are? If I remove orientation = "y", I get a lot of vertical lines! — tnt
– tnt, Commented Mar 22, 2024 at 18:38
@MrFlick, hmmm... ok, I'm trying to show what values fall within the CIs, so if I can't use the quartiles, how should I do this? — tnt
– tnt, Commented Mar 22, 2024 at 18:39
If you look at the answer I've just posted, you'll see that I do! ;-) — Limey
– Limey, Commented Mar 22, 2024 at 18:41

Limey · Accepted Answer · 2024-03-27 15:48:12Z

1

This

dat |>
  ggplot(aes(x = mass, col = sex)) +
  geom_density() +
  stat_summary(
    aes(xintercept = after_stat(x), y = 0),
    fun = mean, geom = "vline", orientation = "y"
  ) +
  geom_vline(
    data = dat %>% 
      group_by(site, sex) %>% 
      reframe(
        tibble(
          value = quantile(mass, probs = c(0.025, 0.975)), 
          prob = c(0.025, 0.975)
        )
      ),
    aes(xintercept = value, col = sex),
    linetype = "dashed"
  ) +
  facet_wrap(~site)

gives

Personally, I find overriding the data argument of a geom_xxx call more intuitive than calling stat_summary and after_stat.

This code gives you what (I think) you expect your code to do. But I support MrFlick's warning about whether it's appropriate.

edited Mar 27, 2024 at 15:48

answered Mar 22, 2024 at 18:35

Limey

12.9k2 gold badges17 silver badges42 bronze badges

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

1 Comment

tnt Over a year ago

Thanks @Limey. I'm still holding out hope for an example for how to do this within stat_summary.

Collectives™ on Stack Overflow

Add quantiles to geom_density plot with stat_summary

1 Answer 1

1 Comment

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Linked

Related