ggplot geom_bar with stat = "sum"

Question

I want to plot a bar chart summing a variable along two dimensions, one will be spread along x, and the other will be spread vertically (stacked).

I would expect the two following instructions to do the same, but they don't and only the 2nd one gives the desired output (where I aggregate the data myself).

I'd like to understand what's going on in the first case, and if there's a way to use ggplot2 's built-in aggregation features to get the right output.

library(ggplot2)
library(dplyr)
p1 <- ggplot(diamonds,aes(cut,price,fill=color)) + 
  geom_bar(stat="sum",na.rm=TRUE)

yielding this plot:

p2 <- ggplot(diamonds %>%
                group_by(cut,color) %>%
                summarize_at("price",sum,na.rm=T),
              aes(cut,price,fill=color)) +
  geom_bar(stat="identity",na.rm=TRUE)

yielding this picture:

Here's where the top of our bars should be, p1 doesn't give these values:

diamonds %>% group_by(cut) %>% summarize_at("price",sum,na.rm=TRUE)
# # A tibble: 5 x 2
# cut    price
# <ord>    <int>
# 1      Fair  7017600
# 2      Good 19275009
# 3 Very Good 48107623
# 4   Premium 63221498
# 5     Ideal 74513487

Answer 1

You might be misunderstanding the stat option for geom_bar. In this case, since you want the values for each factor to be summed up within each bar, and the bars to be colored based off how much of that total sum is in each color, you can simplify the call to geom_col which uses the values as heights for the bar; and therefore "sums" all the values within each category. For example, the following will give the desired output:

p1 <- ggplot(diamonds,aes(cut,price,fill=color)) + 
        geom_col(na.rm=TRUE)

Alternatively, if you want to use geom_bar with a stat call, then you want to use the "identity" stat:

p1 <- ggplot(diamonds,aes(cut,price,fill=color)) + 
        geom_bar(stat = "identity", na.rm=TRUE)

For more information, consider this thread: https://stackoverflow.com/a/27965637/6722506