基于ggplot2在同图绘制Bland-Altman图及多变量平滑曲线

Hey there! It sounds like you’ve nailed the two scatter groups in your Bland-Altman plot, but hitting a snag when adding smooth curves for your pre.moy/post.moy (x-axis) and pre.dif/post.dif (y-axis) pairs. The most common culprit here is working with wide-format data instead of long-format, which confuses ggplot’s variable mapping for the curves. Let’s walk through a solution step by step.

Step 1: Reshape Your Data to Long Format #

First, we need to tidy your data so that the pre/post groups are stored in a single column, and their corresponding mean (moy) and difference (dif) values live in consistent columns. We’ll use tidyr::pivot_longer for this:

library(tidyverse)

# Replace `your_data` with the name of your actual data frame
tidy_data <- your_data %>%
  # Reshape mean values into long format
  pivot_longer(
    cols = c(pre.moy, post.moy),
    names_to = "time_point",
    values_to = "mean_value",
    names_pattern = "(pre|post)\\.moy"  # Extract pre/post from column names
  ) %>%
  # Reshape difference values into long format
  pivot_longer(
    cols = c(pre.dif, post.dif),
    names_to = "time_point_dif",
    values_to = "difference",
    names_pattern = "(pre|post)\\.dif"
  ) %>%
  # Ensure pre/post pairs are matched correctly
  filter(time_point == time_point_dif) %>%
  select(-time_point_dif)  # Remove redundant column

This gives you a clean long-format data frame where each row represents one observation, with a time_point column to distinguish pre vs post, mean_value for your x-axis, and difference for your y-axis.

Step 2: Plot with Scatter Points + Smooth Curves #

Now you can use this tidy data to build your plot. ggplot will automatically handle grouping the scatter points and smooth curves by the time_point variable:

ggplot(tidy_data, aes(x = mean_value, y = difference, color = time_point)) +
  # Add scatter points with slight transparency to avoid overcrowding
  geom_point(alpha = 0.6) +
  # Add smooth loess curves (adjust method to "lm" if you want linear fits)
  geom_smooth(method = "loess", se = FALSE, linewidth = 1) +
  # Customize labels and theme
  labs(
    x = "Mean of Measurements",
    y = "Difference Between Measurements",
    color = "Time Point",
    title = "Bland-Altman Plot with Grouped Smooth Curves"
  ) +
  theme_minimal()

Why This Works #

By using long-format data, you’re letting ggplot handle the grouping logic instead of manually trying to map separate pre/post variables across layers. This avoids the "variable splicing" errors you were seeing, because all relevant data is mapped consistently in a single aes() call.

Troubleshooting Tips #

• If you still get errors, double-check that your original pre.moy/post.moy and pre.dif/post.dif columns have the same number of rows (no missing values that might break the pairing).
• If you prefer linear smooth curves instead of loess, change method = "loess" to method = "lm" in geom_smooth().

内容的提问来源于stack exchange，提问作者PHNM