如何用ggplot和ggpubr为水平森林图添加外部计算的P值？

问题重现 #

你尝试绘制展示比值比（Odds Ratio）的水平森林图，使用ggpubr::stat_pvalue_manual添加P值时触发以下错误：

Error: Discrete value supplied to continuous scale

数据集 #

order<-data.frame(
  Order=factor(c(1:6),levels=c(6:1)),
  OR=c(1,0.92,0.84,0.83,0.81,0.82),
  CI_low=c(NA,0.87,0.79,0.77,0.74,0.74),
  CI_high=c(NA,0.99,0.90,0.89,0.88,0.90),
  p_between=c(NA,0.01550,0.00610,0.72685,0.61190,0.86702),
  p_between_sig=c(NA,"*","**","NS","NS","NS"),
  p_overall=1.662490e-09,
  p_overall_sig=c("****")
)

stat.test<-data.frame(
                group1=c(1,2,3,4,5,1),
                 group2=c(2,3,4,5,6,6),
                 p=c(0.02,0.01,0.73,0.61,0.87,1.6e-09),
                 p.signif=c("*","**","NS","NS","NS","****"))

初始森林图代码 #

library(ggplot2)
library(ggpubr)
fp <- order %>%
  ggplot(aes(y=Order, x=OR, label=Order)) +
  geom_point(size=4, shape=15) +
  theme_classic()+
  geom_errorbarh(aes(xmin=CI_low, xmax=CI_high), height=0)+
  geom_vline(xintercept=1, linetype='longdash') +
  coord_cartesian(xlim=c(.7, 1.25),ylim=c(1,7))+
  labs(x="Odds Ratio",y="Order")+
  annotate("text", x = .85, y = 7, label = "Lower Detection") +
  annotate("text", x = 1.1, y = 7, label = "Higher Detection")
fp

报错的P值添加代码 #

fp +
  stat_pvalue_manual(stat.test, label = "p.signif", tip.length=0.01,
                     y.position = c(1.05,1.05,1.05,1.05,1.05,1.1),
                     coord.flip=T)

错误原因 #

1. coord.flip=T参数冲突：你的初始图已经通过geom_errorbarh实现水平布局，y轴是离散因子（Order），x轴是连续数值（OR）。stat_pvalue_manual的coord.flip=T会强制翻转坐标轴，导致离散值被传入连续刻度，触发错误。
2. 分组不匹配：stat.test中的group1/group2是数值，而order$Order是因子，两者层级对应关系未对齐。
3. y.position设置错误：原y.position数值对应错误，未匹配Order因子的实际位置（因子Order=6对应y=1，Order=1对应y=6）。

修复方案 #

步骤1：修正stat.test的分组 #

将group1和group2转换为与order$Order一致的因子，确保分组匹配：

stat.test <- stat.test %>%
  mutate(
    group1 = factor(group1, levels = c(6:1)),
    group2 = factor(group2, levels = c(6:1))
  )

步骤2：调整stat_pvalue_manual参数 #

• 移除coord.flip=T参数
• 修正y.position为对应因子的位置
• 自定义label参数实现带括号的P值格式

完整修复代码 #

library(ggplot2)
library(ggpubr)
library(dplyr)

# 修正stat.test的分组
stat.test <- stat.test %>%
  mutate(
    group1 = factor(group1, levels = c(6:1)),
    group2 = factor(group2, levels = c(6:1))
  )

# 绘制森林图并添加带括号的P值
fp <- order %>%
  ggplot(aes(y=Order, x=OR, label=Order)) +
  geom_point(size=4, shape=15) +
  theme_classic()+
  geom_errorbarh(aes(xmin=CI_low, xmax=CI_high), height=0)+
  geom_vline(xintercept=1, linetype='longdash') +
  coord_cartesian(xlim=c(.7, 1.3),ylim=c(1,7))+
  labs(x="Odds Ratio",y="Order")+
  annotate("text", x = .85, y = 7, label = "Lower Detection") +
  annotate("text", x = 1.1, y = 7, label = "Higher Detection") +
  # 添加带括号的P值
  stat_pvalue_manual(
    stat.test,
    label = "p = {round(p, 3)} ({p.signif})", # 自定义带括号的P值格式
    tip.length = 0.01,
    y.position = c(1.2, 2.2, 3.2, 4.2, 5.2, 6.5), # 对应每个对比的y位置
    x.position = 1.18 # 设置P值的x位置
  )

fp

效果说明 #

• 修正后的代码会在每个森林图条目的右侧添加带括号的P值和显著性标记
• 整体对比的P值会显示在最上方的位置
• 完全避免了坐标轴翻转带来的离散/连续刻度冲突问题

内容的提问来源于stack exchange，提问作者Hong