ggplot中geom_bar的stat=identity参数含义及代码疑问

Hey there! Let's unpack this confusion step by step—your observations are totally on point, and I’ll break down exactly what’s happening with that code, plus clarify what stat="identity" actually does.

First, what your current code is doing #

Your guess is correct: this code is drawing a bar for every single row in your dataset, not aggregating values for Group A and B. Here’s why:

• When you set stat="identity" in geom_bar(), you’re telling ggplot: "Don’t calculate any new values for the y-axis—use the exact numbers from my Value column as the height of each bar."
• Your data has multiple rows for Group A and Group B (e.g., two A rows with values 20 and 24, two B rows with 74 and 50). Since all these rows share the same x=Group value (A or B), their bars get drawn in the same position.
• You added position="dodge", but since your fill is also mapped to Group, ggplot doesn’t see a reason to dodge the bars (they’re all part of the same fill group). So the bars stack/overlap instead of side-by-side. The tallest bar in each group ends up on top, making it look like only the maximum value is displayed.

What stat="identity" really means #

To put it simply:

• The default behavior of geom_bar() (without specifying stat) is stat="count": it counts how many times each x category appears, and uses that count as the bar height (no y variable needed).
• stat="identity" overrides this—it tells ggplot to use your provided y values directly. This is useful when you’ve already pre-calculated aggregated values (like sums, means, or medians) and just want to plot them as bars.

How to fix it to plot aggregated stats (like mean/median) #

Your hunch that the original goal was to plot summary statistics is spot-on. There are two easy ways to do this:

1. Pre-aggregate your data first (using dplyr)

Calculate the summary statistic you want (e.g., mean) before plotting:

library(dplyr)
# Calculate mean Value for each Group
summary_data <- data %>%
  group_by(Group) %>%
  summarize(Average_Value = mean(Value))

# Now plot the aggregated data
ggplot(summary_data, aes(x = Group, y = Average_Value, fill = Group)) +
  geom_bar(stat="identity", position = "dodge", lwd = 1)

2. Let ggplot handle the aggregation (using stat_summary)

You can skip pre-aggregating and have ggplot calculate the summary on the fly:

ggplot(data, aes(x = Group, y = Value, fill = Group)) +
  stat_summary(fun = mean, geom = "bar", position = "dodge", lwd = 1)

Replace fun = mean with fun = median or fun = sum if you want a different summary statistic.

内容的提问来源于stack exchange，提问作者Agustin