如何在ggplot箱线图中添加格式为“p = ___”的组间比较p值

Hey Simon, let's fix that p-value formatting issue for your boxplots! I’ve run into this exact problem before, so here are a few reliable solutions depending on how you’re currently calculating and adding your p-values:

If you’re using ggsignif to add significance brackets, you can directly customize the annotation text with glue using the annotation parameter. This lets you format p-values exactly as "p = 0.XXXX" instead of just raw numbers.

library(ggplot2)
library(ggsignif)
library(glue)

# Sample dataset (replace with your data)
set.seed(123)
df <- data.frame(
  Group = rep(c("Control", "Treatment1", "Treatment2"), each = 30),
  Value = c(rnorm(30, 10), rnorm(30, 12), rnorm(30, 15))
)

# Create boxplot with formatted p-values
ggplot(df, aes(x = Group, y = Value)) +
  geom_boxplot(fill = "lightblue") +
  geom_signif(
    comparisons = list(c("Control", "Treatment1"), c("Control", "Treatment2")),
    map_signif_level = FALSE,  # Disable star symbols, show raw p-values
    textsize = 4,
    # Custom annotation function to format p-values
    annotation = function(p_val) glue("p = {round(p_val, 4)}")
  )

The annotation argument takes a function that transforms the raw p-value into your desired string format. Adjust round(p_val, 4) to change the number of decimal places as needed.

If you prefer ggpubr for statistical comparisons, use the label.func parameter in stat_compare_means to format your p-values:

library(ggpubr)

ggplot(df, aes(x = Group, y = Value)) +
  geom_boxplot(fill = "lightgreen") +
  stat_compare_means(
    comparisons = list(c("Control", "Treatment1"), c("Control", "Treatment2")),
    method = "t.test",  # Specify your statistical test here
    label.func = function(p_val) glue("p = {round(p_val, 4)}"),
    size = 4
  )

This overrides the default label format and lets you inject the "p = " prefix directly into each annotation.

If you’re calculating p-values manually (e.g., with t.test, anova, etc.), pre-format the p-value strings in your data frame first, then use geom_text to add them to the plot:

# Step 1: Calculate p-values manually
p_results <- data.frame(
  group_pair = c("Control vs Treatment1", "Control vs Treatment2"),
  p_val = c(
    t.test(df$Value[df$Group == "Control"], df$Value[df$Group == "Treatment1"])$p.value,
    t.test(df$Value[df$Group == "Control"], df$Value[df$Group == "Treatment2"])$p.value
  ),
  # Pre-format p-values as "p = XXXX" strings
  p_label = glue("p = {round(p_val, 4)}")
)

# Step 2: Define positions for labels (adjust based on your plot's axis)
p_results$x_pos <- c(1.5, 2)  # Midpoints between group positions (1=Control, 2=Treatment1, 3=Treatment2)
p_results$y_pos <- max(df$Value) + 1  # Place labels above the tallest boxplot

# Step 3: Add formatted labels to the plot
ggplot(df, aes(x = Group, y = Value)) +
  geom_boxplot(fill = "pink") +
  geom_text(
    data = p_results,
    aes(x = x_pos, y = y_pos, label = p_label),
    size = 4,
    fontface = "bold"
  )

The key here is pre-generating the p_label column as strings—this avoids issues with ggplot trying to map numeric values directly to text labels.

Common Pitfalls to Avoid #

• Don’t try to use glue directly inside geom_text(aes(label = ...)) without pre-processing: ggplot’s aesthetic mapping expects a column name or transformed value, so pre-formatting the string in your data frame is safer.
• If you’re seeing raw numbers instead of formatted text, double-check that your label column is a character string (not numeric). Use as.character() if needed.

Hope one of these methods works for your project! Let me know if you need help adapting it to your specific dataset or statistical tests.

内容的提问来源于stack exchange，提问作者simon_p