如何在R语言DataFrame中保留某变量最常见的前N个观测值？

原始数据 #

dt <- structure(list(ID = 1:10, City = c("New York", "New York", "LA", 
"LA", "LA", "Boston", "Chicago ", "New York", "LA", "New York"
), Random_Info_To_Keep = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L)), class = "data.frame", row.names = c(NA, -10L))

方法一：Base R 实现 #

1. 统计各城市的出现次数
2. 筛选出出现次数Top2的城市
3. 保留包含这些城市的行

# 统计城市出现频次
city_freq <- table(dt$City)
# 获取Top2高频城市
top2_cities <- names(sort(city_freq, decreasing = TRUE)[1:2])
# 筛选目标行
dt2 <- dt[dt$City %in% top2_cities, ]

方法二：dplyr 实现（tidyverse风格） #

借助dplyr的计数和筛选功能，代码更简洁直观：

library(dplyr)

dt2 <- dt %>%
  add_count(City) %>%  # 给每行添加对应城市的出现次数
  filter(n %in% sort(unique(n), decreasing = TRUE)[1:2]) %>%  # 筛选Top2频次的城市行
  select(-n)  # 移除临时生成的计数列

验证结果 #

执行dput(dt2)会得到预期输出：

structure(list(ID = c(1L, 2L, 3L, 4L, 5L, 8L, 9L, 10L), City = c("New York", 
"New York", "LA", "LA", "LA", "New York", "LA", "New York"), 
    Random_Info_To_Keep = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L)), class = "data.frame", row.names = c(NA, 
-8L))

内容的提问来源于stack exchange，提问作者Jamie