R语言ggplot2自定义图例需求：为不同中心节点设置区分色图例

Hey there! I totally get where you're coming from—ggplot2's legend system can feel way more convoluted than what you might be used to in Python when you're just starting out with R. Let's get your centers showing up in the legend with different colors properly.

What Was Off in Your Original Code #

• You hardcoded colors (color = "blue"/color = "red") directly in geom_point instead of mapping them through aes(), so ggplot didn't link those points to the legend.
• Default point shapes (like shape 19) don't support fill colors, so using fill= in your aes wasn't doing anything visible.
• The ellipses and points weren't sharing a unified legend mapping, leading to inconsistent or missing legend entries.

Revised Code #

library(ggplot2)
library(MASS)
library(car)
library(robustbase)
data("airquality")

# Select only Ozone and Temp variables
air = airquality[c("Ozone" , "Temp")]
air = na.omit(air)

# Mahalanobis calculations
air.center = colMeans(air)
air.cov = cov(air)
rad = sqrt(qchisq(p = 0.95 , df = ncol(air)))
ellipse <- as.data.frame(ellipse(center = air.center , shape = air.cov , radius = rad , segments = 150 , draw = FALSE))
colnames(ellipse) <- colnames(air)
distances <- mahalanobis(x = air , center = air.center , cov = air.cov)
cutoff <- qchisq(p = 0.95 , df = ncol(air))

# MCD calculations
Y_mcd <- covMcd(air)
ellipse_mcd <- data.frame(ellipse(center = Y_mcd$center, shape = Y_mcd$cov, radius= rad, segments=100,draw=FALSE))
colnames(ellipse_mcd) <- colnames(air)

# Modified plot code
plot_fig <- ggplot(air , aes(x = Ozone , y = Temp)) + 
  # Base scatter points
  geom_point(size = 2) + 
  # Mahalanobis ellipse: map fill/color to a label, disable inherited aes to avoid conflicts
  geom_polygon(data = ellipse, aes(fill = "Mahalanobis", color = "Mahalanobis"), 
               alpha = 0.5, inherit.aes = FALSE) + 
  # Mahalanobis center: use shape 21 (supports fill) and map to label
  geom_point(aes(x = air.center[1], y = air.center[2], fill = "Mahalanobis", color = "Mahalanobis"), 
             size = 5, shape = 21) + 
  # Label outliers
  geom_text(data=subset(air, distances > cutoff), 
            aes(Ozone,Temp,label=row.names(air[distances > cutoff ,])), 
            hjust = 1 , vjust = -1.5 ,size = 3.5)+ 
  # Axis labels and title
  ylab("Temperature Values") + 
  xlab("Ozone Values") +
  ggtitle("Mahalanobis Distance vs. MCD") + 
  # MCD ellipse: map fill/color to label
  geom_polygon(data=ellipse_mcd,aes(fill = "MCD", color = "MCD"), 
               alpha=0.3, inherit.aes = FALSE) + 
  # MCD center: shape 21 for fill mapping
  geom_point(aes(x = Y_mcd$center[1], y = Y_mcd$center[2], fill = "MCD", color = "MCD"), 
             size = 6, shape = 21) +
  # Manual scales to set colors and unify legend
  scale_fill_manual(name = "Method", 
                    values = c("Mahalanobis" = "blue", "MCD" = "red")) +
  scale_color_manual(name = "Method", 
                     values = c("Mahalanobis" = "blue", "MCD" = "red")) +
  # Legend position adjustment
  theme( legend.position = c(0.95, 0.15), legend.justification = c("right", "top") )

plot_fig

Key Changes Explained #

1. Shape Selection: We use shape = 21 for the center points—this is a circle with a separate fill (inner color) and color (border color), which lets the legend properly display the point style.
2. Aesthetic Mapping: All elements (ellipses and centers) have their fill and color mapped to a label ("Mahalanobis" or "MCD") inside aes(), instead of hardcoding colors. This tells ggplot to create legend entries for these labels.
3. Unified Legend: We use scale_fill_manual and scale_color_manual with the same legend name ("Method") and color values. This merges the ellipse and point entries into a single, clean legend.
4. inherit.aes = FALSE: Added to geom_polygon to prevent inheriting unnecessary aesthetic mappings from the main ggplot call, avoiding conflicts.

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