ggplot2中用stat_function绘制局部填充正态曲线及x变量使用疑问

Hey there! Let's break down how ggplot2 handles the x variable when plotting normal curves—especially in no-data scenarios and when adding shaded regions. This is such a common sticking point for new R users, so you’re definitely not alone in this confusion!

First, remember that ggplot2 is built around data frames, but when you start with ggplot() (no data supplied), stat_function steps in to create an implicit sequence of x values for you. By default, this sequence spans from -3 to 3 (a standard range for normal distributions), but you can override this by:

• Setting xlim() to define your desired x-axis range (e.g., xlim(mean - 3*sd, mean + 3*sd) to cover 3 standard deviations around your mean)
• Manually creating a data frame of x values (more explicit, which I recommend for clarity)

Let’s say you want a normal curve with mean = 5 and standard deviation = 2. Here are two solid approaches:

Approach 1: Use stat_function with args #

This is the quick no-data method—ggplot handles generating x values for you:

library(ggplot2)

ggplot() +
  stat_function(fun = dnorm, 
                args = list(mean = 5, sd = 2),  # Pass your mean/sd to dnorm
                color = "navy") +
  xlim(5 - 3*2, 5 + 3*2) +  # Lock x-axis to mean ±3 SD
  labs(x = "Value", y = "Probability Density")

Approach 2: Manual x Data Frame (More Transparent) #

If you want full control over the x values, create a data frame explicitly:

# Generate 1000 evenly spaced x values across your desired range
x_vals <- seq(5 - 3*2, 5 + 3*2, length.out = 1000)
# Calculate density for each x
norm_df <- data.frame(
  x = x_vals,
  density = dnorm(x_vals, mean = 5, sd = 2)
)

ggplot(norm_df, aes(x = x, y = density)) +
  geom_line(color = "navy") +
  labs(x = "Value", y = "Probability Density")

Your funcShaded function works because stat_function passes every generated x value to it one by one. The core idea is: return the normal density for x values in your target region, and return 0 (or NA) for values outside—this tells geom="area" to only fill the area where the function returns a non-zero value.

Example funcShaded Implementation #

Let’s say you want to shade the region between x=3 and x=7 (for our mean=5, sd=2 curve):

funcShaded <- function(x) {
  # Define your target interval
  lower_bound <- 3
  upper_bound <- 7
  # Calculate normal density for x
  dens <- dnorm(x, mean = 5, sd = 2)
  # Return density only if x is in the interval; else return 0
  ifelse(x >= lower_bound & x <= upper_bound, dens, 0)
}

Add the Shaded Layer to Your Plot #

ggplot() +
  # Base normal curve
  stat_function(fun = dnorm, args = list(mean = 5, sd = 2), color = "navy") +
  # Shaded region
  stat_function(fun = funcShaded, 
                geom = "area", 
                fill = "#84CA72", 
                alpha = 0.3) +  # Alpha makes the fill transparent
  xlim(5 - 3*2, 5 + 3*2) +
  labs(x = "Value", y = "Probability Density")

Pro Tip: Make funcShaded Reusable #

Instead of hardcoding mean, sd, or bounds, turn them into function parameters and pass them via args—this way you can reuse the function for different curves:

funcShaded <- function(x, mean, sd, lower, upper) {
  dens <- dnorm(x, mean = mean, sd = sd)
  ifelse(x >= lower & x <= upper, dens, 0)
}

# Use it with custom parameters
ggplot() +
  stat_function(fun = dnorm, args = list(mean = 5, sd = 2), color = "navy") +
  stat_function(fun = funcShaded,
                args = list(mean = 5, sd = 2, lower = 3, upper = 7),
                geom = "area", fill = "#84CA72", alpha = 0.3) +
  xlim(5 - 3*2, 5 + 3*2)

• Implicit vs Explicit x: In no-data mode, stat_function generates x values automatically. For full control, create your own x data frame.
• Function Requirements: Any function passed to stat_function must accept x as its first parameter—ggplot will feed every x value into it to compute the corresponding y value.
• Shading Logic: Shading works by returning non-zero density values only for your target x range; geom="area" fills the area under these non-zero points.

内容的提问来源于stack exchange，提问作者chainhomelow