如何在存在联系时总结数据集中的前 3 个最高值

Question

我有一个数据框 (my_data)，我想只计算 3 个最高值的总和，即使可能存在联系。我是 R 的新手，我用过 dplyr。

A tibble: 15 x 3
   city      month number
   <chr>     <chr>  <dbl>
 1 Lund      jan       12
 2 Lund      feb       12
 3 Lund      mar       18
 4 Lund      apr       28
 5 Lund      may       28
 6 Stockholm jan       15
 7 Stockholm feb       15
 8 Stockholm mar       30
 9 Stockholm apr       30
10 Stockholm may       10
11 Uppsala   jan       22
12 Uppsala   feb       30
13 Uppsala   mar       40
14 Uppsala   apr       60
15 Uppsala   may       30

这是我试过的代码：

# For each city, count the top 3 of variable number
my_data %>% group_by(city) %>% top_n(3, number) %>% summarise(top_nr = sum(number))

预期（想要）的输出是：

# A tibble: 3 x 2
  city      top_nr
  <chr>      <dbl>
1 Lund          86
2 Stockholm     75
3 Uppsala      130

但实际的 R 输出是：

# A tibble: 3 x 2
  city      top_nr
  <chr>      <dbl>
1 Lund          86
2 Stockholm     90
3 Uppsala      160

似乎如果有联系，所有联系的值都包含在总和中。我只想计算 3 个具有最高值的唯一实例。

如有任何帮助，我们将不胜感激！ :)

Answer 1

我们可以做一个distinct来删除重复的元素。 top_n 的工作方式是，如果值重复，它将保留那么多重复行

my_data %>% 
   distinct(city, number, .keep_all = TRUE) %>%
   group_by(city) %>%
   top_n(3, number) %>%
   summarise(top_nr = sum(number))

更新

根据OP的新输出，在top_n输出后（不是arranged），得到'number'降序排列得到sum 的前 3 个 'number'

my_data %>% 
   group_by(city) %>% 
   top_n(3, number) %>% 
   arrange(city,  desc(number)) %>% 
   summarise(number = sum(head(number, 3)))
# A tibble: 3 x 2
#  city      number
#  <chr>      <int>
#1 Lund          74
#2 Stockholm     75
#3 Uppsala      130

数据

my_data <- structure(list(city = c("Lund", "Lund", "Lund", "Lund", "Lund", 
"Stockholm", "Stockholm", "Stockholm", "Stockholm", "Stockholm", 
"Uppsala", "Uppsala", "Uppsala", "Uppsala", "Uppsala"), month = c("jan", 
"feb", "mar", "apr", "may", "jan", "feb", "mar", "apr", "may", 
"jan", "feb", "mar", "apr", "may"), number = c(12L, 12L, 18L, 
28L, 28L, 15L, 15L, 30L, 30L, 10L, 22L, 30L, 40L, 60L, 30L)), 
class = "data.frame", row.names = c("1", 
"2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", 
"14", "15"))

Answer 2

这个tidyverse（实际上，dplyr）解决方案几乎等于，但是filter是数据帧而不是top_n。

library(tidyverse)

my_data %>%
  group_by(city) %>%
  arrange(desc(number), .by_group = TRUE) %>%
  filter(row_number() %in% 1:3) %>%
  summarise(top_nr = sum(number))
## A tibble: 3 x 2
#  city      top_nr
#  <chr>      <int>
#1 Lund          74
#2 Stockholm     75
#3 Uppsala      130

Answer 3

如果没有 top_n()，生活可能会更简单：

dat %>%
  group_by(city) %>%
  summarize(
    top_nr = sum(tail(sort(number), 3))
    )

如何在存在联系时总结数据集中的前 3 个最高值

How to summarize the top 3 highest values in a dataset when there are ties

r

dplyr

top-n

更新

数据