Merge legend from different geom and aes?

Question

the idea is simple:
(1)Use red dot for actual data;
(2)Use blue dashed line for fitted value.
But I just could not find a way to make such legend.

My fig and code below: fig

library(ggplot)
x <- seq(1,100,1)
y <- 2*x + runif(100,-5,5)
fit_y <- 2*x

df <- data.frame(x=x, y=y,fit_y=fit_y)
df

ggplot(df)+
  geom_point(aes(x=x, y=y, color='point', shape='point'))+
  geom_line(aes(x=x, y=fit_y, color='line', linetype='line')) +
  scale_color_manual(values=c('point'='red', 'line'='blue')) +
  scale_linetype_manual(values = c('line'='longdash')) +
  scale_shape_manual(values = c('point'=1 ))

And I am pretty confused about that: Why is color legend composed of line and point?
Now it is a "blue line + blue point" PLUS "red line + red point".
But I only assign red to geom_point and blue to geom_line! Why is this?

Try: ggplot(df) + geom_point(aes(x=x, y=y), color='blue') + geom_line(aes(x=x, y=fit_y), color='red', linetype='longdash') ? — zx8754
No legend showing this way...(Maybe because no attributes in aes() mapping) — Jiaqi Ding

teunbrand teunbrand · Accepted Answer · 2021-05-17T15:09:01

Perhaps the best way is to not bother mapping aesthetics for shape and linetype at all, and instead overriding the display of the colour legend directly. This is also the suggested workaround for guides indiscriminately applying glyphs for every key.

library(ggplot2)
x <- seq(1,100,1)
y <- 2*x + runif(100,-5,5)
fit_y <- 2*x

df <- data.frame(x=x, y=y,fit_y=fit_y)

ggplot(df)+
  geom_point(aes(x=x, y=y, color='point'),
             shape = 1)+
  geom_line(aes(x=x, y=fit_y, color='line', linetype='line'),
            linetype = 'longdash') +
  scale_color_manual(
    values=c('point'='red', 'line'='blue'),
    guide = guide_legend(
      override.aes = list(
        shape = c(1, NA),
        linetype = c(NA, 5)
      )
    )
  )

^{Created on 2021-05-17 by the reprex package (v1.0.0)}

Merge legend from different geom and aes?

3 Answers