如何在ggplot中合并多组中心化z分数变量生成森林图（免bind_rows）

核心思路 #

使用tidyr::pivot_longer()将宽格式的因变量（或其回归结果）转换为长格式，替代手动调用bind_rows合并多个数据集的冗余操作，再结合ggplot统一绘制森林图。

示例实现 #

1. 准备数据与批量回归 #

先生成包含自变量和多个中心化因变量的数据集，再批量拟合回归并提取结果（替代手动逐个拟合后合并）：

library(tidyverse)
library(broom)

# 生成原始数据
set.seed(123)
df <- tibble(
  gender = sample(c("Male", "Female"), 100, replace = TRUE),
  age = rnorm(100, 30, 5),
  zscore_math = rnorm(100, 0, 1),
  zscore_science = rnorm(100, 0, 1)
)

# 批量拟合回归并提取结果，自动合并为长格式
reg_results <- df %>%
  select(starts_with("zscore")) %>%
  imap_df(function(y, outcome_name) {
    lm(y ~ gender + age, data = df) %>%
      tidy(conf.int = TRUE) %>%
      mutate(outcome = outcome_name)
  })

2. 绘制多因变量森林图 #

基于自动生成的长格式回归结果，直接绘制森林图：

ggplot(reg_results, aes(x = estimate, y = term, color = outcome)) +
  # 添加零值参考线
  geom_vline(xintercept = 0, linetype = "dashed", color = "gray50") +
  # 绘制系数点
  geom_point(position = position_dodge(width = 0.6), size = 2) +
  # 绘制置信区间
  geom_errorbarh(aes(xmin = conf.low, xmax = conf.high),
                 position = position_dodge(width = 0.6), height = 0.2) +
  # 设置标签与主题
  labs(x = "标准化系数估计值", y = "自变量", color = "因变量") +
  theme_minimal()

替代方案：处理宽格式系数表 #

如果你的数据是宽格式（比如已手动计算好每个因变量的系数和置信区间），用pivot_longer快速转长：

# 示例宽格式系数表
coef_wide <- tibble(
  term = c("genderFemale", "age"),
  math_est = c(0.2, 0.1),
  math_low = c(-0.1, -0.05),
  math_high = c(0.5, 0.25),
  science_est = c(0.3, 0.08),
  science_low = c(0.05, -0.07),
  science_high = c(0.55, 0.23)
)

# 转换为长格式
coef_long <- coef_wide %>%
  pivot_longer(
    cols = -term,
    names_to = c("outcome", ".value"),
    names_sep = "_"
  )

# 绘图代码和上述一致
ggplot(coef_long, aes(x = est, y = term, color = outcome)) +
  geom_vline(xintercept = 0, linetype = "dashed", color = "gray50") +
  geom_point(position = position_dodge(width = 0.6), size = 2) +
  geom_errorbarh(aes(xmin = low, xmax = high),
                 position = position_dodge(width = 0.6), height = 0.2) +
  labs(x = "系数估计值", y = "自变量", color = "因变量") +
  theme_minimal()

关键优势 #

• 避免了手动bind_rows重复拼接多个单变量结果的冗余代码，减少出错概率。
• pivot_longer和imap_df的组合能高效处理多因变量的批量操作，代码更简洁易维护。
• 用position_dodge实现不同因变量结果的并排展示，森林图可读性更强。

内容的提问来源于stack exchange，提问作者Luis