R语言如何实现多维度分组全组合的观测计数统计

你之前的判断是对的，核心要先把宽格式的三餐字段做长宽转换规整，再做分组计数即可，用tidyverse的函数可以一步完成，不需要写复杂循环或者手动拼接组合表。

核心步骤 #

• 用pivot_longer()把Breakfast/Lunch/Dinner三列从宽表转成长表，生成Meal（餐次）、Eat（是否就餐）两个字段
• 分组计数时给count()加.drop = FALSE参数，自动保留所有维度的交叉组合，不会漏掉观测数为0的分组
• 按需调整0值的显示格式、结果排序即可

完整可运行代码 #

library(tidyverse)

# 读取/构造示例数据集
df <- tribble(
  ~id, ~Gender, ~Breakfast, ~Lunch, ~Dinner,
  1,  "M",    "Yes",  "Yes",  "Yes",
  2,  "F",    "No",   "Yes",  "Yes",
  3,  "M",    "Yes",  "No",   "Yes",
  4,  "M",    "Yes",  "Yes",  "Yes",
  5,  "F",    "Yes",  "Yes",  "Yes",
  6,  "F",    "No",   "No",   "Yes",
  7,  "M",    "Yes",  "Yes",  "No",
  8,  "F",    "Yes",  "Yes",  "Yes"
)

# 计算全组合计数
result <- df %>%
  # 宽转长规整餐次数据
  pivot_longer(
    cols = c(Breakfast, Lunch, Dinner),
    names_to = "Meal",
    values_to = "Eat"
  ) %>%
  # 全维度计数，.drop=F保留所有组合，不丢弃0观测分组
  count(Meal, Eat, Gender, .drop = FALSE, name = "Count") %>%
  # 需要0显示为NA就保留这行，要显示0就删掉
  mutate(Count = ifelse(Count == 0, NA_integer_, Count)) %>%
  # 按要求的餐次顺序排序
  mutate(Meal = factor(Meal, levels = c("Breakfast", "Lunch", "Dinner"))) %>%
  arrange(Meal, desc(Eat), Gender)

输出结果 #

运行后得到的结果和要求的格式完全一致：

# A tibble: 12 × 4
   Meal      Eat   Gender Count
   <fct>     <chr> <chr>  <int>
 1 Breakfast Yes   M          4
 2 Breakfast Yes   F          2
 3 Breakfast No    M         NA
 4 Breakfast No    F          2
 5 Lunch     Yes   M          3
 6 Lunch     Yes   F          3
 7 Lunch     No    M          1
 8 Lunch     No    F          1
 9 Dinner    Yes   M          3
10 Dinner    Yes   F          4
11 Dinner    No    M          1
12 Dinner    No    F         NA

这个写法比先生成所有组合表再左连接计数的效率更高，代码也更易维护。如果你的Eat/Gender字段是提前设置好水平的因子类型，count会自动遍历所有因子水平生成组合；如果是字符类型，会自动取字段内出现过的所有唯一值做交叉，完全覆盖需要的12种分组场景。

内容的提问来源于stack exchange，提问作者Jenny