使用 dplyr 进行 group_by + 汇总时保留其他列

Question

I want to do a group_by + summarise operation on only two columns with one group attribute while keeping the other three columns unchanged which have the same number for every row. How can I do that? eg

> data<- data.frame(a=1:10, b=rep(1,10), c=rep(2,10), d=rep(3,10), e= c("small", "med", "larg", "larg", "larg", "med", "small", "small", "small", "med"))
> data %>% group_by(e) %>% summarise(a=mean(a))
# A tibble: 3 × 2
  e         a
  <chr> <dbl>
1 larg   4   
2 med    6   
3 small  6.25

but I want

# A tibble: 3 × 5
  e         a b     c     d
  <chr> <dbl> <dbl> <dbl> <dbl>
1 larg   4    1     2     3
2 med    6    1     2     3
3 small  6.25 1     2     3

group_by + summarise always drops other columns. How can I do that?

Answer 1

Add the other columns to group_by :

> library(tidyverse)
> data <- data.frame(a=1:10, b=rep(1,10), c=rep(2,10), d=rep(3,10), e= c("small", "med", "larg", "larg", "larg", "med", "small", "small", "small", "med"))
> data %>% group_by(e, b, c, d) %>% summarise(a=mean(a))
`summarise()` has grouped output by 'e', 'b', 'c'. You can override using the `.groups` argument.
# A tibble: 3 x 5
# Groups:   e, b, c [3]
  e         b     c     d     a
  <chr> <dbl> <dbl> <dbl> <dbl>
1 larg      1     2     3  4   
2 med       1     2     3  6   
3 small     1     2     3  6.25

Answer 2

And you can always calculate a new variable with group + summarise and keep the rest of your dataframe "intact" adding across() in the summarise. This could be useful if your other variables arent going to be the same always.

data %>% group_by(e) %>% 
    summarise(a=mean(a), across())

    # A tibble: 10 x 5
# Groups:   e [3]
   e         a     b     c     d
   <chr> <dbl> <dbl> <dbl> <dbl>
 1 larg   4        1     2     3
 2 larg   4        1     2     3
 3 larg   4        1     2     3
 4 med    6        1     2     3
 5 med    6        1     2     3
 6 med    6        1     2     3
 7 small  6.25     1     2     3
 8 small  6.25     1     2     3
 9 small  6.25     1     2     3
10 small  6.25     1     2     3

Answer 3

It is unclear how many columns you want to treat as grouping variable. If the number is small, @tauft's answer is sufficient. Otherwise, we can use across with group_by so that we can use <tidy-select> to select the columns to group.

library(dplyr)

data2 <- data %>%
  group_by(across(-a)) %>%
  summarise(a = mean(a), .groups = "drop") %>%
  relocate(e, a, .before = b)
data2
# # A tibble: 3 x 5
#   e         a     b     c     d
#   <chr> <dbl> <dbl> <dbl> <dbl>
# 1 larg   4        1     2     3
# 2 med    6        1     2     3
# 3 small  6.25     1     2     3

The above can also written as follows.

data2 <- data %>%
  group_by(across(b:e)) %>%
  summarise(a = mean(a), .groups = "drop") %>%
  relocate(e, a, .before = b)