求助：R-Studio中OLS多元回归模型系数对比图绘制问题

Hey there! Let's get that side-by-side coefficient comparison plot working for you. The core issue here is that ggplot thrives on tidy (long-form) data, and it sounds like your combined model results might be in a wide format that doesn't let ggplot distinguish between the two models. Here's a step-by-step fix:

Step 1: Tidy Your Model Results #

First, use the broom package to extract coefficients from each model separately, add a label for each model, then combine them into one tidy data frame. This makes it easy for ggplot to group and color by model.

# Load required packages
library(broom)
library(dplyr)
library(ggplot2)

# Assume your two OLS models are named model1 and model2 (replace with your actual model names)
tidy_model1 <- tidy(model1) %>% mutate(model = "Model 1")  # Add model identifier
tidy_model2 <- tidy(model2) %>% mutate(model = "Model 2")

# Combine the tidy results into one data frame
combined_tidy <- bind_rows(tidy_model1, tidy_model2)

Step 2: Plot with ggplot (Color-Coded & Side-by-Side) #

Now use the tidy data frame to build your plot. We'll use color = model to distinguish the two models, and position_dodge() to keep their coefficients side by side instead of overlapping.

ggplot(combined_tidy, aes(x = term, y = estimate, color = model, group = model)) +
  # Add coefficient points, dodged to avoid overlap
  geom_point(position = position_dodge(width = 0.5), size = 2) +
  # Add 95% confidence intervals (adjust multiplier if you want different CI)
  geom_errorbar(aes(ymin = estimate - 1.96 * std.error, ymax = estimate + 1.96 * std.error),
                width = 0.2, position = position_dodge(width = 0.5)) +
  # Customize labels and theme
  labs(title = "OLS Coefficient Comparison: Model 1 vs Model 2",
       x = "Predictor Variables",
       y = "Coefficient Estimate",
       color = "Model") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))  # Rotate x-labels for readability

If You Already Have a Wide-Format Combined Data Frame #

If you've already merged your models into a wide format (e.g., columns like estimate_model1, estimate_model2, std.error_model1), convert it to long-form first with tidyr::pivot_longer():

library(tidyr)

# Convert wide to long format
combined_long <- your_wide_data %>%
  pivot_longer(
    cols = starts_with("estimate"),
    names_to = "model",
    values_to = "estimate",
    names_prefix = "estimate_"  # Remove the prefix from model names
  ) %>%
  pivot_longer(
    cols = starts_with("std.error"),
    names_to = "model_error",
    values_to = "std.error",
    names_prefix = "std.error_"
  ) %>%
  filter(model == model_error) %>%  # Match each model to its standard error
  select(-model_error)

Then use combined_long in the ggplot code above instead of combined_tidy.

This should give you a clear, color-coded plot where you can easily compare coefficients across both models. Let me know if you need to tweak anything for your specific dataset!

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