How to create a theoretical conditional error distribution plot using ggplot2

tidyverse
1

Probably a rather naive question but I am really having trouble finding an answer or recalling any solution from my reading of ggplot2 related books.
I want to plot the conditional error distribution in a simple linear regression model with same or different error variance.
Something like,

image
image638×618 19.5 KB

I know how to plot the usual distribution using x and y axis, but could it be rotated like that of the above picture?
Thanks!

2

Here a sketch of how to make a plot similar to your image.

library(tidyverse)
DF <- data.frame(Z = rep(seq(-4,4,0.02),3),
                 Y = -c(dnorm(seq(-4,4,0.02),-1,sd = 0.5)-1,
                       dnorm(seq(-4,4,0.02),mean = 0,sd = 1.5),
                       dnorm(seq(-4,4,0.02),mean = 0.5, sd = 1)+1),
                 Theta = rep(c(1,0,-1),each = 401))
DF2 <- DF |> filter(Z > -2.0)
ggplot(DF,aes(Z,Y,group = Theta)) + geom_line() + 
  geom_hline(yintercept = c(-1,0,1)) + 
  geom_vline(xintercept = -2) +
  geom_ribbon(aes(ymax = Theta, ymin = Y), fill = "grey80", data = DF2) + 
  coord_flip() +
  theme_bw()

Created on 2024-11-08 with reprex v2.1.1

1 Like
3

Thanks! This solution is really elegant and brilliant!