Filter top n largest groups in data.frame

Question

For the example data:

set.seed(2222)
example_data <- data.frame(col1 = 1:15,
                           col2 = 16:30, 
                           group = sample(1:3, 15, replace = TRUE))

   col1 col2 group
1     1   16     2
2     2   17     1
3     3   18     3
4     4   19     2
5     5   20     3
6     6   21     1
7     7   22     3
8     8   23     1
9     9   24     3
10   10   25     1
11   11   26     2
12   12   27     2
13   13   28     2
14   14   29     3
15   15   30     3

I want to find the top n groups with the most number of records.

Let's say I want to get the top 2 groups with the most number of records. In the data, this would be group 3 and 2:

example_data %>% 
  group_by(group) %>% 
  summarise(n = n())

# A tibble: 3 x 2
  group     n
  <int> <int>
1     1     4
2     2     5
3     3     6

The expected output is:

   col1 col2 group
1     1   16     2
2     3   18     3
3     4   19     2
4     5   20     3
5     7   22     3
6     9   24     3
7    11   26     2
8    12   27     2
9    13   28     2
10   14   29     3
11   15   30     3

There are very similar questions with very similar answers [here](https://stackoverflow.com/q/43110349/5325862) and [here](https://stackoverflow.com/q/52532080/5325862) — camille, May 29 '19 at 15:20
similar, but one looks for the top n within a group and the other one filters for groups of a fixed size. — clemens, May 29 '19 at 15:29

Ronak Shah · Accepted Answer · 2019-05-29T14:46:08.840

We can use table to calculate frequency for each group, sort them in decreasing order, subset the top 2 entries and filter the respective groups.

library(dplyr)

example_data %>%
   filter(group %in% names(sort(table(group), decreasing = TRUE)[1:2]))


#   col1 col2 group
#1     1   16     2
#2     3   18     3
#3     4   19     2
#4     5   20     3
#5     7   22     3
#6     9   24     3
#7    11   26     2
#8    12   27     2
#9    13   28     2
#10   14   29     3
#11   15   30     3

Also you can directly use this in base R subset

subset(example_data, group %in% names(sort(table(group), decreasing = TRUE)[1:2]))

akrun · Answer 2 · 2019-05-29T15:22:22.810

We can use tidyverse methods for this. Create a frequency column with add_count, arrange by that column and filter the rows where the 'group' is in the last two unique 'group' values

library(dplyr)
example_data %>% 
   add_count(group) %>% 
   arrange(n) %>%
   filter(group %in% tail(unique(group), 2)) %>%
   select(-n)
# A tibble: 11 x 3
#    col1  col2 group
#  <int> <int> <int>
# 1     1    16     2
# 2     4    19     2
# 3    11    26     2
# 4    12    27     2
# 5    13    28     2
# 6     3    18     3
# 7     5    20     3
# 8     7    22     3
# 9     9    24     3
#10    14    29     3
#11    15    30     3

Or using data.table

library(data.table)
setDT(example_data)[group %in% example_data[, .N, group][order(-N), head(group, 2)]]

tmfmnk · Answer 3 · 2019-05-30T12:52:04.917

With dplyr, you can also do:

example_data %>%
 add_count(group) %>%
 filter(dense_rank(desc(n)) <= 2) %>%
 select(-n)

   col1  col2 group
   <int> <int> <int>
 1     1    16     2
 2     3    18     3
 3     4    19     2
 4     5    20     3
 5     7    22     3
 6     9    24     3
 7    11    26     2
 8    12    27     2
 9    13    28     2
10    14    29     3
11    15    30     3

Or:

example_data %>%
 add_count(group) %>%
 slice(which(dense_rank(desc(n)) <= 2)) %>%
 select(-n)

score 2 · Answer 4 · answered May 29 '19 at 14:51

2

Another dplyr method could be:

example_data %>% 
  group_by(group) %>% 
  count() %>% 
  ungroup() %>% 
  top_n(n = 2, wt = n) %>% 
  select(-n) %>% 
  inner_join(example_data)

answered May 29 '19 at 14:51

OTStats

1,820
1
13
22

utubun · Answer 5 · 2019-05-29T15:20:12.977

Yet another solution (similar to @OTStats):

library(dplyr)

example_data %>% 
  inner_join(
    count(., group) %>% top_n(2, n) %>% dplyr::select(group)
    )

#    col1 col2 group
# 1     1   16     2
# 2     3   18     3
# 3     4   19     2
# 4     5   20     3
# 5     7   22     3
# 6     9   24     3
# 7    11   26     2
# 8    12   27     2
# 9    13   28     2
# 10   14   29     3
# 11   15   30     3

Filter top n largest groups in data.frame

5 Answers5