如何使用dplyr的group_by函数合并两个字符型列？

给定数据框df1：

df1 <- tribble(
  ~Country, ~Gender, ~var1, ~var2, ~var3, ~Income,
  "Bangladesh", "F", 2.5, 3, 1.5, "LM",
  "Bangladesh", "M", 4.5, 4.3, 2.7, "LM",
  "Laos", "F", 2.7, 3.2, 6.5, "LM", 
  "Laos", "M", 3.5, 5.1, 8.2, "LM", 
  "Ghana", "F", 8.5, 5, 7.5, "LM",
  "Ghana", "M", 4, 6.7, 1.3, "LM",
  "China", "F", 4.3, 6.1, 2.5, "UM",
  "China", "M", 6.2, 2.8, 6.8, "UM",
)

可以用group_by合并数值列：

df1 %>% 
  group_by(Country, subgroup = var1 + var2) %>%
  summarise()

但尝试用+合并字符型列时触发错误：

df1 %>% 
  group_by(Country, subgroup = Gender + Income) %>%
  summarise()

#Error: ! non-numeric argument to binary operator

期望生成类似df2的结构：

df2 <- tribble(
  ~Country, ~subgroup, 
  "Bangladesh", "F", 
  "Bangladesh", "M", 
  "Laos", "F",  
  "Laos", "M", 
  "Ghana", "F", 
  "Ghana", "M", 
  "China", "F", 
  "China", "M",
  "Bangladesh", "LM", 
  "Bangladesh", "LM", 
  "Laos", "LM",  
  "Laos", "LM", 
  "Ghana", "LM", 
  "Ghana", "LM", 
  "China", "UM", 
  "China", "UM",
)

解决方案 #

字符型列不能用+运算，且你的目标是将Gender和Income的内容分别作为subgroup的行，需要先将数据转为长格式，再按需处理：

1. 生成目标结构df2 #

使用tidyr::pivot_longer将两列转为单列：

library(dplyr)
library(tidyr)

df_result <- df1 %>%
  select(Country, Gender, Income) %>% # 保留需要的列
  pivot_longer(
    cols = c(Gender, Income), # 指定要转换的列
    names_to = NULL, # 丢弃原列名
    values_to = "subgroup" # 新列名
  )

# 输出结果
df_result

2. 按Country和subgroup分组汇总 #

如果需要对分组后的数据做统计（比如计数）：

df_result %>%
  group_by(Country, subgroup) %>%
  summarise(count = n(), .groups = "drop")

补充：若需拼接字符生成subgroup #

如果你的需求是将Gender和Income拼接成单个字符串（如"F_LM"），用paste()实现：

df1 %>%
  group_by(Country, subgroup = paste(Gender, Income, sep = "_")) %>%
  summarise(mean_var1 = mean(var1), .groups = "drop")

内容的提问来源于stack exchange，提问作者rez