如何在ggplot2中保持变量顺序与输入一致？

尝试使用以下R代码绘制模型回归系数的森林图：

library(ggplot2)
library(tidyverse)

# Create a data.frame
forestPlot <- tribble(
  ~factor, ~PR, ~Low, ~High,
  #------/------/------/------/
  "Breastfeeding: <17 weeks vs 26-51", 0.75,0.52,1.10,
  "17-25 weeks vs 26-51", 0.81, 0.50, 1.31,
  "≥52 weeks vs 26-51", 1.20, 0.88, 1.63,
  
  "Mother's age", 0.99, 0.96, 1.02,
  
  "Mother's education: High school vs tertiary", 1.64, 1.14, 2.36,
  "Vocational vs tertiary",1.31, 0.92, 1.86,
  
  "Mother's country of birth: UK/Ireland vs AU/NZ", 0.53, 0.17, 1.63,
  "India vs AU/NZ",1.66, 1.05, 2.61,
  "Asia-except India vs AU/NZ",1.86,1.28,2.69,
  "Other vs AU/NZ",0.83, 0.43, 1.58,
  
  "IRSAD: 1 vs 5", 1.48, 0.77, 2.83,
  "2 vs 5", 1.47, 0.79, 2.78,
  "3 vs 5", 1.09, 0.56, 2.14,
  "4 vs 5", 1.36, 0.72, 2.56,
)
forestPlot

# Factoring variables of the model
forestPlot$factor <- factor(forestPlot$factor, levels = c(
  "Breastfeeding: <17 weeks vs 26-51", "17-25 weeks vs 26-51", "≥52 weeks vs 26-51", 
  "Mother's age", "Mother's education: High school vs tertiary", 
  "Vocational vs tertiary", "Mother's country of birth: UK/Ireland vs AU/NZ", 
  "India vs AU/NZ", "Asia-except India vs AU/NZ", "Other vs AU/NZ", "IRSAD: 1 vs 5", 
  "2 vs 5", "3 vs 5", "4 vs 5"))
forestPlot

# Make a forest plot
library(forcats)
Model1A <- forestPlot %>%
  ggplot(aes(x = fct_inorder(factor), y = PR, ymin = Low, ymax = High)) +
  geom_errorbar(width=.2, size = 1.0, show.legend = F) +
  geom_point(size= 3.5, shape=21, fill="white", show.legend = F)+ 
  geom_hline(yintercept = 1, linetype = 2, col = "red", size = 0.8) +
  coord_flip() +
  labs(title = "A. Model for detal caries prevalance", subtitle = "Breastfeeding as main exposure",
       x = "", y = "PR (95% CI)") +
  theme(
    text = element_text(family = "Georgia"),
    plot.title = element_text(size = 14, color = "grey10", face = "bold"),
    plot.subtitle = element_text(face = "italic", color = "gray20", size = 12),
    plot.caption = element_text(face = "italic", size = 12, color = "gray40"),
    axis.title.x = element_text(face = "bold", size = 12, color = "grey20"),
    axis.title.y = element_blank(),
    panel.grid.major = element_line(color = "gray90"),
    panel.grid.minor.y = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.background = element_rect(fill = "#fcfbfd"))
Model1A

但绘图中预测变量的顺序从底部开始显示，希望保持与输入变量相同的顺序（从顶部开始），使用fct_inorder(factor)函数并未生效，求解决方法。

问题根源在于coord_flip()会翻转坐标轴的显示顺序：你已经手动指定了factor的levels为输入的从上到下顺序，但翻转后这个顺序会被反转，导致第一个变量出现在图的底部。

只需对factor变量反转因子顺序即可修正，有两种简单方法：

方法1：在aes中使用fct_rev() #

将ggplot代码中的x = fct_inorder(factor)替换为x = fct_rev(factor)，利用你已经提前设置好的levels反转顺序，配合coord_flip()就能让输入的第一个变量显示在图的顶部：

Model1A <- forestPlot %>%
  ggplot(aes(x = fct_rev(factor), y = PR, ymin = Low, ymax = High)) +
  geom_errorbar(width=.2, size = 1.0, show.legend = F) +
  geom_point(size= 3.5, shape=21, fill="white", show.legend = F)+ 
  geom_hline(yintercept = 1, linetype = 2, col = "red", size = 0.8) +
  coord_flip() +
  labs(title = "A. Model for detal caries prevalance", subtitle = "Breastfeeding as main exposure",
       x = "", y = "PR (95% CI)") +
  theme(
    text = element_text(family = "Georgia"),
    plot.title = element_text(size = 14, color = "grey10", face = "bold"),
    plot.subtitle = element_text(face = "italic", color = "gray20", size = 12),
    plot.caption = element_text(face = "italic", size = 12, color = "gray40"),
    axis.title.x = element_text(face = "bold", size = 12, color = "grey20"),
    axis.title.y = element_blank(),
    panel.grid.major = element_line(color = "gray90"),
    panel.grid.minor.y = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.background = element_rect(fill = "#fcfbfd"))
Model1A

方法2：设置坐标轴limits #

不修改aes，而是添加scale_x_discrete(limits = rev(levels(forestPlot$factor)))来指定x轴的显示顺序，同样能达到效果：

Model1A <- forestPlot %>%
  ggplot(aes(x = factor, y = PR, ymin = Low, ymax = High)) +
  geom_errorbar(width=.2, size = 1.0, show.legend = F) +
  geom_point(size= 3.5, shape=21, fill="white", show.legend = F)+ 
  geom_hline(yintercept = 1, linetype = 2, col = "red", size = 0.8) +
  coord_flip() +
  scale_x_discrete(limits = rev(levels(forestPlot$factor))) + # 添加这一行
  labs(title = "A. Model for detal caries prevalance", subtitle = "Breastfeeding as main exposure",
       x = "", y = "PR (95% CI)") +
  theme(
    text = element_text(family = "Georgia"),
    plot.title = element_text(size = 14, color = "grey10", face = "bold"),
    plot.subtitle = element_text(face = "italic", color = "gray20", size = 12),
    plot.caption = element_text(face = "italic", size = 12, color = "gray40"),
    axis.title.x = element_text(face = "bold", size = 12, color = "grey20"),
    axis.title.y = element_blank(),
    panel.grid.major = element_line(color = "gray90"),
    panel.grid.minor.y = element_blank(),
    panel.grid.minor.x = element_blank(),
    panel.background = element_rect(fill = "#fcfbfd"))
Model1A

两种方法都能让变量按照你输入的顺序从图的顶部开始显示，无需额外调整factor的levels设置。

内容的提问来源于stack exchange，提问作者user332276