Stacked bar graph with proportion on y axis instead of count

Question

I'm using the gss_cat tibble from forcats (tidyverse), and the first thing I decided to plot was race x income. So I did a bar plot, but it turns out the total of surveyed people is not the same for each race:

gss_cat %>%
  filter(!rincome %in% c("No answer", "Don't know", "Not applicable")) %>%
  ggplot(aes(x = race, fill = rincome))+
  geom_bar(position = "stack")+
  #coord_polar(theta = "y")+
  theme_classic()+
  scale_fill_viridis_d(direction = -1)

The same happens with a pie chart (so without that # at coord_polar): So I thought I do y = ..prop.., which theoretically would use the proportion instead of the total, but this seems to make everything the same:

gss_cat %>%
  filter(!rincome %in% c("No answer", "Don't know", "Not applicable")) %>%
  ggplot(aes(x = race, y = ..prop.., fill = rincome))+
  geom_bar(position = "stack")+
  #coord_polar(theta = "y")+
  theme_classic()+
  scale_fill_viridis_d(direction = -1)

I then tried doing group = race, which in my experience works for geom_count, but it didn't for this:

gss_cat %>%
  filter(!rincome %in% c("No answer", "Don't know", "Not applicable")) %>%
  ggplot(aes(x = race, y = ..prop.., fill = rincome, group = race))+
  geom_bar(position = "stack")+
  #coord_polar(theta = "y")+
  theme_classic()+
  scale_fill_viridis_d(direction = -1)

It just threw the fill out the window entirely. There was no warning.

Why is this happening and how can I change it? I didn't find anything on R Documentation, the ggplot2 homepage, this website – which is one of my favorites... I don't know where to go from here.

Answer 1

Try with position=fill:

library(forcats)
library(ggplot2)
library(dplyr)
#Code
gss_cat %>%
  filter(!rincome %in% c("No answer", "Don't know", "Not applicable")) %>%
  ggplot(aes(x = race, fill = rincome))+
  geom_bar(position = "fill")+
  #coord_polar(theta = "y")+
  theme_classic()+
  scale_fill_viridis_d(direction = -1)

Output: