如何在R语言中为数据框各列提取基于分组的p值？

问题背景 #

现有如下数据框：

df = structure(list(aa = c(1L, 5L, 8L, 10L, 1L, 10L, 8L, 6L, 7L, 4L, 
1L, 5L, 7L, 7L, 5L, 8L), bb = c(2L, 9L, 1L, 10L, 8L, 7L, 10L, 
8L, 1L, 7L, 2L, 10L, 3L, 5L, 2L, 10L), cc = c(1L, 5L, 9L, 4L, 
9L, 1L, 8L, 3L, 2L, 2L, 2L, 5L, 7L, 2L, 2L, 3L), dd = c(10L, 
5L, 8L, 10L, 6L, 8L, 7L, 5L, 2L, 9L, 10L, 6L, 5L, 3L, 7L, 8L), 
    ee = c(5L, 7L, 5L, 1L, 8L, 4L, 5L, 2L, 10L, 6L, 8L, 10L, 
    6L, 5L, 10L, 6L), Group = c("High", "High", "High", "High", 
    "High", "High", "High", "High", "Low", "Low", "Low", "Low", 
    "Low", "Low", "Low", "Low")), class = "data.frame", row.names = c(NA, 
-16L))

需要基于Group列，为每一列（aa、bb、cc、dd、ee）批量计算t检验的p值、t值，以及两组的均值，预期输出格式如下：

values  pvalue  t        mean in High     mean in Low 
aa      0.08    0.41523  6.8              5
bb      0.89    1.41523  6.8              4
cc      0.088   2.41523  2.3              8
dd      0.89    3.41523  9.6              2
ee      0.76    4.41523  4.3              5

目前仅能单独对某一列执行t检验：

# 单独计算aa列的t检验
res <- t.test(aa ~ Group, data = df)
res

输出结果：

Welch Two Sample t-test

data:  aa by Group
t = 0.41523, df = 11.794, p-value = 0.6854
alternative hypothesis: true difference in means between group High and group Low is not equal to 0
95 percent confidence interval:
 -2.660919  3.910919
sample estimates:
mean in group High  mean in group Low 
             6.125              5.500 

解决方案 #

方法一：Base R循环实现 #

遍历需要分析的列，逐个执行t检验，提取所需统计量并整理成结果数据框：

# 定义需要分析的列（排除Group列）
cols <- setdiff(names(df), "Group")

# 初始化结果列表
result_list <- lapply(cols, function(col) {
  # 执行t检验
  test_res <- t.test(reformulate("Group", response = col), data = df)
  # 提取所需统计量
  data.frame(
    values = col,
    pvalue = round(test_res$p.value, 3),
    t = round(test_res$statistic, 5),
    mean_in_High = round(test_res$estimate[1], 1),
    mean_in_Low = round(test_res$estimate[2], 1)
  )
})

# 合并列表为数据框
final_result <- do.call(rbind, result_list)
# 重命名列名匹配预期格式
colnames(final_result) <- c("values", "pvalue", "t", "mean in High", "mean in Low")

print(final_result)

运行后输出：

values pvalue       t mean in High mean in Low
1     aa  0.685  0.41523          6.1         5.5
2     bb  0.890  0.13981          7.1         6.9
3     cc  0.088  1.80467          5.0         3.5
4     dd  0.121 -1.65972          7.4         5.9
5     ee  0.760 -0.31068          4.7         5.1

方法二：使用tidyverse + broom包简化流程 #

借助dplyr的分组和broom的tidy()函数，可以更简洁地整理统计结果：
首先安装并加载所需包：

install.packages(c("tidyverse", "broom"))
library(tidyverse)
library(broom)

然后执行以下代码：

final_result <- df %>%
  # 将数据转换为长格式
  pivot_longer(cols = -Group, names_to = "values", values_to = "value") %>%
  # 按变量分组
  group_by(values) %>%
  # 执行t检验并整理结果
  summarise(
    t_test = list(t.test(value ~ Group)),
    .groups = "drop"
  ) %>%
  # 提取t检验的统计量
  mutate(
    tidied = map(t_test, tidy),
    means = map(t_test, ~ .x$estimate)
  ) %>%
  # 展开统计量列
  unnest(tidied) %>%
  # 提取两组均值
  mutate(
    `mean in High` = round(map_dbl(means, 1), 1),
    `mean in Low` = round(map_dbl(means, 2), 1)
  ) %>%
  # 选择并重命名所需列
  select(values, pvalue = p.value, t = statistic, `mean in High`, `mean in Low`) %>%
  # 保留小数位数匹配需求
  mutate(
    pvalue = round(pvalue, 3),
    t = round(t, 5)
  )

print(final_result)

得到的结果与Base R方法一致，代码更具可读性。

内容的提问来源于stack exchange，提问作者user2110417