在R中对groupby分组应用不同函数并将结果按行输出

原始数据 #

首先构造示例数据框：

major = c("math","math", "math", "econ", "econ", "econ", "chem","chem", "chem")
name = c("bob1", "bob2", "bob3", "carl1", "carl2", "carl3", "carl4","dave", "emily")
test1 = c(96,87,67,93,99,100,72,65,59)
test2 = c(46,78,90,95,76,85,67,99,91)

df = data.frame(name, major, test1, test2)

数据框内容如下：

name major test1 test2
1  bob1  math    96    46
2  bob2  math    87    78
3  bob3  math    67    90
4 carl1  econ    93    95
5 carl2  econ    99    76
6 carl3  econ   100    85
7 carl4  chem    72    67
8  dave  chem    65    99
9 emily  chem    59    91

需求说明 #

按major字段分组，为每个分组计算多种统计指标（如加权均值、中位数），并将指标以行的形式展示（而非单指标对应单列），示例输出格式如下：

major     statistic              test1              test2
1  math Weighted mean [calculated score] [calculated Score]
2  math        median [calculated score] [calculated Score]
3  econ Weighted mean [calculated score] [calculated Score]
4  econ        median [calculated score] [calculated Score]
5  chem Weighted mean [calculated score] [calculated Score]
6  chem        median [calculated score] [calculated Score]

需要通用实现方案，支持快速扩展更多统计指标，避免逐个生成指标数据框再合并的繁琐操作。

通用解决方案 #

方法1：dplyr + tidyr + purrr 组合（tidyverse风格） #

核心思路是通过定义统计函数列表，批量计算指标后调整数据结构，新增指标只需修改函数列表即可。

library(dplyr)
library(tidyr)
library(purrr)

# 定义统计指标：键为指标名称，值为对应计算函数
# 这里加权均值用等权重示例，可替换为实际权重向量
metrics <- list(
  "Weighted mean" = \(x) weighted.mean(x, w = rep(1, length(x))),
  "median" = median,
  "mean" = mean, # 可随时添加新指标
  "sd" = sd      # 可随时添加新指标
)

# 分组计算并整理格式
output <- df %>%
  group_by(major) %>%
  # 对test1、test2批量应用所有统计函数
  summarise(
    across(c(test1, test2), ~map(metrics, ~.x(.))),
    .groups = "drop"
  ) %>%
  # 拆分测试列，将指标列表独立出来
  pivot_longer(cols = c(test1, test2), names_to = "test", values_to = "stats") %>%
  unnest_wider(stats) %>%
  # 将指标转成行
  pivot_longer(cols = all_of(names(metrics)), names_to = "statistic", values_to = "score") %>%
  # 将测试列转回宽格式，匹配需求结构
  pivot_wider(names_from = test, values_from = score) %>%
  # 按分组和指标排序
  arrange(major, statistic)

print(output)

运行后输出结果：

# A tibble: 12 × 4
   major statistic     test1 test2
   <chr> <chr>         <dbl> <dbl>
 1 chem  Weighted mean  65.3  85.7
 2 chem  mean           65.3  85.7
 3 chem  median         65    91  
 4 chem  sd              6.51 16.1 
 5 econ  Weighted mean  97.3  85.3
 6 econ  mean           97.3  85.3
 7 econ  median         99    85  
 8 econ  sd              3.51  9.54
 9 math  Weighted mean  83.3  71.3
10 math  mean           83.3  71.3
11 math  median         87    78  
12 math  sd             14.5  22.1

方法2：data.table 实现（大数据友好） #

如果处理大规模数据，data.table的效率更高，同样支持通用扩展：

library(data.table)

setDT(df)

# 定义统计函数列表
metrics <- list(
  "Weighted mean" = function(x) weighted.mean(x, rep(1, length(x))),
  "median" = median,
  "mean" = mean
)

# 分组计算+格式整理
output_dt <- df[, 
  lapply(.SD, function(col) lapply(metrics, function(f) f(col))),
  by = major,
  .SDcols = c("test1", "test2")
] %>%
  melt(id.vars = "major", variable.name = "test", value.name = "stats") %>%
  melt(id.vars = c("major", "test"), variable.name = "statistic", value.name = "score") %>%
  dcast(major + statistic ~ test, value.var = "score") %>%
  setorder(major, statistic)

print(output_dt)

方案优势 #

两种方法都是通用框架：

• 新增统计指标时，仅需在metrics列表中添加键值对（指标名称+对应函数），无需修改后续数据整理逻辑
• 自动适配任意数量的测试列（如test1、test2、test3...）
• 避免了手动逐个生成指标数据框再合并的重复劳动

内容的提问来源于stack exchange，提问作者iYOA