Hi Mattia, awesome first question on Stack Overflow! I’ve got a practical R solution to help you overlay those two correlation plots exactly how you described—showing both correlation values in the lower triangle and comparing ellipses in the upper triangle. Let’s dive in:

Step 1: Prepare Your Environment & Data #

First, make sure you’ve got the corrplot package installed and loaded. We’ll start with sample data (swap this out with your actual corr.pre and corr.post matrices):

# Install if needed: install.packages("corrplot")
library(corrplot)

# Create sample pre/post correlation matrices (replace with your data!)
set.seed(123)
corr.pre <- cor(matrix(rnorm(50), ncol = 5))
corr.post <- cor(matrix(rnorm(50, mean = 0.2), ncol = 5))

Step 2: Build the Combined Plot #

We’ll layer two plots together: first the post-period ellipses, then overlay the pre-period ellipses in a distinct color. Finally, we’ll add both correlation values to the lower triangle:

# 1. Base plot: Post-period ellipses in upper triangle, blank lower triangle
corrplot.mixed(
  corr.post,
  upper = "ellipse",
  lower = "blank",
  upper.col = colorRampPalette(c("#2c3e50", "#ffffff", "#e74c3c"))(100),
  tl.pos = "lt",  # Position variable labels at top-left
  tl.cex = 0.8,   # Adjust label size if needed
  cl.pos = "r"    # Keep color bar on right
)

# 2. Overlay pre-period ellipses in a darker, distinct color
corrplot(
  corr.pre,
  add = TRUE,
  type = "upper",
  method = "ellipse",
  col = colorRampPalette(c("#16a085", "#ffffff", "#c0392b"))(100),
  bg = "transparent",
  tl.pos = "n",  # Hide duplicate labels
  cl.pos = "n"   # Hide duplicate color bar
)

# 3. Add both pre/post correlation values to the lower triangle
n <- nrow(corr.pre)
for (i in 2:n) {
  for (j in 1:(i - 1)) {
    # Calculate position coordinates (matches corrplot's grid)
    x <- j
    y <- n - i + 1
    
    # Add pre-period value (bottom of cell, dark teal)
    text(x, y - 0.18, round(corr.pre[i, j], 2), cex = 0.7, col = "#16a085")
    # Add post-period value (top of cell, dark red)
    text(x, y + 0.18, round(corr.post[i, j], 2), cex = 0.7, col = "#c0392b")
  }
}

# 4. Add legend to clarify pre/post periods
legend(
  "bottomright",
  legend = c("Pre-period", "Post-period"),
  col = c("#16a085", "#c0392b"),
  lwd = 3,
  cex = 0.8,
  bty = "n"  # Remove legend box for cleaner look
)

Key Tips for Customization #

• Match Matrix Dimensions: Double-check that corr.pre and corr.post have the same number of variables in the same order—otherwise the overlay will be misaligned!
• Adjust Colors: Swap out the hex codes for any color palette that works better for your data (try RColorBrewer palettes for professional options).
• Tweak Text: Modify round() to change decimal places, or adjust cex to make the correlation values larger/smaller.
• Label Placement: If variable names are too long, use tl.rot = 45 to rotate labels in the base plot.

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