How can we use multiple geom_smooth in a single chart?

Question

I would like to use multiple geom_smooth layers in a single ggplot2 chart. When I try to do that, the color scheme gets screwed up. Here is an example demonstrating what is happening.

We construct a simple dataframe we want to visualize.

df = data.frame(x = c("a", "b", "c"),
             y1 = seq(1, 3),
             y1_upr = seq(2, 4),
             y1_lwr = seq(0, 2),
             y2 = seq(2, 4),
             y2_upr = seq(2.5, 4.5),
             y2_lwr = seq(1.5, 3.5))

We can visualize y1 and y2 easily.

plot_obj = ggplot(data = df, aes(x = x, group = 1)) + 
  geom_line(aes(y = y1, colour = "y1")) + 
  geom_line(aes(y = y2, colour = "y2")) +     
  scale_colour_manual("", breaks = c("y1", "y2"), values = c("blue", "red"))
plot_obj

If we add one geom_smooth, the behavior is still as expected.

plot_obj + 
  geom_smooth(aes(y = y1, ymin = y1_lwr, ymax = y1_upr), stat="identity", fill="blue", alpha=0.2)

Lastly, we add the second geom_smooth layer.

plot_obj + 
  geom_smooth(aes(y = y1, ymin = y1_lwr, ymax = y1_upr), stat="identity", fill="blue", alpha=0.2) + 
  geom_smooth(aes(y = y2, ymin = y2_lwr, ymax = y2_upr), stat="identity", fill="red", alpha=0.2)

Notice that the top line is no longer red in the last chart. Why is this happening and how can it be fixed? Thank you!

the `melt` answer gives the idea http://stackoverflow.com/questions/3777174/plotting-two-variables-as-lines-using-ggplot2-on-the-same-graph — user20650, Sep 07 '16 at 22:10
`geom_smooth` draws blue lines. So you'd need to use `color = "red"` in the `geom_smooth` layer if you want to add separate layers like this. You would have noticed it in both layers if one of your lines didn't happen to be blue. — aosmith, Sep 07 '16 at 22:16
You can use `color=NA` to turn off the lines drawn by `geom_smooth` which is over-plotting your first lines. Or you can change the order of the layers (smooth first, then line). These are just other ways to solve the problem pointed out by @aosmith. (Although reshaping is a better long term strategy in my opinion.) — MrFlick, Sep 07 '16 at 22:21

aosmith · Accepted Answer · 2019-01-14T15:16:13.587

Certainly reshaping your dataset will make things easier, and is the recommended approach. However, if you want to keep using separate layers:

As you haven't mapped a color for geom_smooth, it uses the default color of blue for the smoothed lines it drew. If you want just the ribbon, use geom_ribbon instead.

ggplot(data = df, aes(x = x, group = 1)) + 
    geom_line(aes(y = y1, colour = "y1")) + 
    geom_line(aes(y = y2, colour = "y2")) +     
    scale_colour_manual("", breaks = c("y1", "y2"), values = c("blue", "red")) +
    geom_ribbon(aes(ymin = y1_lwr, ymax = y1_upr), stat="identity", fill="blue", alpha=0.2) + 
    geom_ribbon(aes(ymin = y2_lwr, ymax = y2_upr), stat="identity", fill="red", alpha=0.2)

Otherwise you'll need to map your colors for each smooth layer within aes or set them manually to red and blue or NA outside of aes.

ggplot(data = df, aes(x = x, group = 1)) + 
    geom_line(aes(y = y1, colour = "y1")) + 
    geom_line(aes(y = y2, colour = "y2")) +     
    scale_colour_manual("", breaks = c("y1", "y2"), values = c("blue", "red")) +
    geom_smooth(aes(y = y1, ymin = y1_lwr, ymax = y1_upr, colour = "y1"), 
              stat="identity", fill="blue", alpha=0.2) + 
    geom_smooth(aes(y = y2, ymin = y2_lwr, ymax = y2_upr, colour = "y2"), 
              stat="identity", fill="red", alpha=0.2)

score 3 · Answer 2 · edited Jan 12 '19 at 16:55

I would probably do something closer to this:

library(dplyr)
df1 <- df %>%
    select(x,contains("y1")) %>%
    rename(y = y1, y_upr = y1_upr, y_lwr = y1_lwr) %>%
    mutate(grp = "y1")
df2 <- df %>%
    select(x,contains("y2")) %>%
    rename(y = y2, y_upr = y2_upr, y_lwr = y2_lwr) %>%
    mutate(grp = "y2")
df_all <- bind_rows(df1, df2)

ggplot(df_all,aes(x = x, y = y, ymin = y_lwr, ymax = y_upr, group = grp)) +
    geom_line(aes(color = grp)) + 
    geom_ribbon(aes(fill = grp), alpha = 0.2)

How can we use multiple geom_smooth in a single chart?

2 Answers2