关于R语言prcomp做PCA绘图：轴百分比显示及指定PC组合绘图的问题

Hey there! Let's tackle your two PCA plotting needs step by step. First, we'll store your PCA result properly to reuse it (instead of calling prcomp() every time), then add variance percentages to your axes, and finally create biplots for any PC pair you want.

Step 1: Store PCA Result & Calculate Variance Explained #

First, let's save your PCA output and compute the variance explained percentage for each principal component. This will make it easy to reference later in plots.

# Replace 'x' with your actual dataset
pca_result <- prcomp(x, scale.=TRUE) # Scale=TRUE is critical for PCA if variables have different units

# Extract variance explained percentages (rounded to 2 decimal places)
var_exp_pct <- round(summary(pca_result)$importance["Proportion of Variance", ] * 100, 2)

Step 2: Add Variance Explained to Plot Axes #

A. Scree Plot (Default PCA Plot) #

The default plot(pca_result) shows eigenvalues (variance) per PC. To add variance percentages as annotations and improve axis labels:

plot(pca_result, type="l", main="Scree Plot", 
     ylab="Variance Explained", xlab="Principal Component")
# Add text labels for variance percentage above each point
text(x=1:length(var_exp_pct), y=pca_result$sdev^2, 
     labels=paste0(var_exp_pct, "%"), pos=3)

B. PCA Scores Scatter Plot #

If you want a scatter plot of PC scores (e.g., PC1 vs PC2) with axis labels showing variance percentages:

# Extract PCA scores
scores <- pca_result$x

# Plot PC1 vs PC2 with custom axis labels
plot(scores[, c("PC1", "PC2")], 
     xlab=paste0("PC1 (", var_exp_pct["PC1"], "% variance)"),
     ylab=paste0("PC2 (", var_exp_pct["PC2"], "% variance)"),
     main="PCA Scores: PC1 vs PC2")

C. Biplot with Variance Explained Labels #

The default biplot() doesn't include variance percentages, but we can pass custom xlab and ylab arguments to fix that:

biplot(pca_result, 
       xlab=paste0("PC1 (", var_exp_pct["PC1"], "% variance)"),
       ylab=paste0("PC2 (", var_exp_pct["PC2"], "% variance)"),
       main="Biplot: PC1 vs PC2")

Step 3: Create Biplots for Specific PC Combinations #

To plot pairs like PC2 vs PC3, use the choices argument in biplot(). This takes a vector of two integers specifying which PCs to use for the x and y axes.

# Biplot for PC2 (x-axis) vs PC3 (y-axis)
biplot(pca_result, 
       choices=c(2, 3), # First number = x-axis PC, second = y-axis PC
       xlab=paste0("PC2 (", var_exp_pct["PC2"], "% variance)"),
       ylab=paste0("PC3 (", var_exp_pct["PC3"], "% variance)"),
       main="Biplot: PC2 vs PC3")

You can swap the numbers to get any combination—for example, choices=c(3,1) would plot PC3 vs PC1.

Full Example with Iris Dataset #

Here's a complete, runnable example using the built-in iris dataset to test everything:

data(iris)
x <- iris[, 1:4] # Use numerical features

# Run PCA
pca_result <- prcomp(x, scale.=TRUE)
var_exp_pct <- round(summary(pca_result)$importance["Proportion of Variance", ] * 100, 2)

# Scree plot with annotations
plot(pca_result, type="l", main="Scree Plot - Iris Dataset",
     ylab="Variance", xlab="Principal Component")
text(x=1:4, y=pca_result$sdev^2, labels=paste0(var_exp_pct, "%"), pos=3)

# PC1 vs PC2 scores plot with species colors
scores <- pca_result$x
plot(scores[, c("PC1", "PC2")], 
     xlab=paste0("PC1 (", var_exp_pct["PC1"], "% variance)"),
     ylab=paste0("PC2 (", var_exp_pct["PC2"], "% variance)"),
     main="PCA Scores: PC1 vs PC2 - Iris",
     col=iris$Species, pch=16)
legend("topright", legend=levels(iris$Species), col=1:3, pch=16)

# PC1 vs PC2 biplot
biplot(pca_result, 
       xlab=paste0("PC1 (", var_exp_pct["PC1"], "% variance)"),
       ylab=paste0("PC2 (", var_exp_pct["PC2"], "% variance)"),
       main="Biplot: PC1 vs PC2 - Iris")

# PC2 vs PC3 biplot
biplot(pca_result, 
       choices=c(2,3),
       xlab=paste0("PC2 (", var_exp_pct["PC2"], "% variance)"),
       ylab=paste0("PC3 (", var_exp_pct["PC3"], "% variance)"),
       main="Biplot: PC2 vs PC3 - Iris")

内容的提问来源于stack exchange，提问作者M. arman