ggplot2中离散变量映射size的警告原因、时机及优化方法咨询

Great question! Let’s unpack this warning and your options clearly.

First: Is this warning a serious red flag? #

Short answer: No, it’s not a critical error that breaks your plot. The warning is ggplot2’s way of nudging you about visual effectiveness, not functionality.

Here’s why it pops up: Human brains are much better at distinguishing discrete attributes like shapes or colors than they are at parsing discrete size differences. If your variable has multiple unordered categories, readers will struggle to quickly map a specific size to a category—especially if the size differences are subtle. It’s a usability warning, not a "your code is broken" alert.

Better Alternatives for Discrete Variables #

If you want your plot to be as readable as possible, here are the recommended approaches:

1. Map to Shape (shape) #

This is the go-to for unordered discrete variables. Different shapes are instantly distinguishable, even for colorblind readers (if you pick shapes with distinct forms, not just filled/unfilled).

Example code:

library(ggplot2)

# Sample data
df <- data.frame(
  x = rnorm(50),
  y = rnorm(50),
  category = sample(c("A", "B", "C"), 50, replace = TRUE)
)

ggplot(df, aes(x = x, y = y, shape = category)) +
  geom_point(size = 3) +  # Set a consistent base size for all points
  scale_shape_manual(values = c(16, 17, 18))  # Customize shapes to your preference

2. Map to Color (color or fill) #

Color is another strong choice for discrete variables. Use color for outline points, fill for filled shapes (like shape = 21). You can use built-in palettes or custom colors to ensure contrast.

Example code:

ggplot(df, aes(x = x, y = y, color = category)) +
  geom_point(size = 3) +
  scale_color_brewer(palette = "Set2")  # Accessible, distinct palette

3. Use Size Only If Your Variable Is Ordered #

If your "discrete" variable actually has an inherent order (e.g., "Small", "Medium", "Large"), you can convert it to an ordered factor to make the size mapping logical—and often suppress the warning.

Example code:

# Convert to ordered factor
df$ordered_category <- factor(df$category, 
                              levels = c("A", "B", "C"), 
                              ordered = TRUE)

ggplot(df, aes(x = x, y = y, size = ordered_category)) +
  geom_point() +
  scale_size_discrete(range = c(2, 6))  # Control the range of sizes

If You Must Use Size for Unordered Discrete Variables #

If you have a specific reason to stick with size (e.g., a requirement from a report), you can manually set sizes to maximize difference and optionally suppress the warning (though this isn’t ideal for readability):

ggplot(df, aes(x = x, y = y, size = category)) +
  geom_point() +
  scale_size_manual(values = c(2, 5, 8))  # Big differences between categories
# To suppress the warning:
suppressWarnings(
  ggplot(df, aes(x = x, y = y, size = category)) +
    geom_point() +
    scale_size_manual(values = c(2, 5, 8))
)

Just keep in mind: This will still be harder for readers to interpret than shape or color.

内容的提问来源于stack exchange，提问作者Pere