R ggplot2 - geom_smooth with gradient color from a

Question

Is there a way to plot a smoothed curve (x=var1, y=var2) and color it with respect to a third continuous variable (z=var3)? I am using the following code:

    library(ggplot2)

    x = runif(100,-20,20)
    y = 2*x+x^2+rnorm(100,0,50)
    z = 0.5*x+rnorm(100,0,2)
    df = data.frame(x=x,y=y,z=z)

    ggplot(data=df,aes(x=x,y=y))+geom_smooth(method='loess', aes(color=z),se=F) 

However, the smoothed line is still solid blue.

Using the internal variable "..y.." instead of var3 colors the line with respect to var2.

    ggplot(data=df,aes(x=x,y=y))+geom_smooth(method='loess', aes(color=..y..),se=F) 

Is there another internal variable to call in order to color the line with respect to var3?

I am able to generate the desired plot with geom_line, but I would like to have it smoothed instead.

Answer 1

You're on the right track using geom_line, you just need to use it on pre-smoothed data. Take your dataframe as above, then:

df$predict <- predict(loess(y~x, data = df))

ggplot(df, aes(x = x,y = predict)) +
  geom_line(aes(colour = z)) 

This can generate ugly results if your x has big gaps; they'll come out as flat segments between points. There are workarounds for that by feeding newdata= to predict() and storing it in a second dataframe, but then you need to also recalculate z for those new x values.