Creating a dynamic Group By

Question

df = data.frame(
  A = c(1, 4, 5, 13, 2),
  B = c("Group 1", "Group 3", "Group 2", "Group 1", "Group 2"),
  C = c("Group 3", "Group 2", "Group 1", "Group 2", "Group 3")
)

df %>%
  group_by(B) %>%
  summarise(val = mean(A))

df %>%
  group_by(C) %>%
  summarise(val = mean(A))

Instead of writing a new chunck of code for each unique set of group_by I would like to create a loop that would iterate through the df data frame and save the results into a list or a data frame.

I would like to see how the average value of feature A is spread acorss features B and C, without having to write a new chunck of code for each categorical feature in the data set.

I tried this:

List_Of_Groups <- map_df(df, function(i) {
  df %>% 
    group_by(!!!syms(names(df)[1:i])) %>% 
    summarize(newValue = mean(A))
})

Answer 1

Using purrr's map, you can apply the chunk of code you specified to all the columns that are character. Basically you map the names of character variables to the function that follows

purrr::map(names(df %>% select(where(is.character))), function(i) {
  df %>% 
    group_by(!!sym(i)) %>% 
    summarize(newValue = mean(A))
})

Output

# [[1]]
# A tibble: 3 x 2
#   B       newValue
#   <chr>      <dbl>
# 1 Group 1      7  
# 2 Group 2      3.5
# 3 Group 3      4  
# 
# [[2]]
# A tibble: 3 x 2
#   C       newValue
#   <chr>      <dbl>
# 1 Group 1      5  
# 2 Group 2      8.5
# 3 Group 3      1.5

Answer 2

You can pivot it long, using A as an identifier, and then group by:

library(tidyr)
df %>% pivot_longer(-A) %>% group_by(name,value) %>% summarize(val=mean(A))
# A tibble: 6 x 3
# Groups:   name [2]
  name  value     val
  <chr> <fct>   <dbl>
1 B     Group 1   7  
2 B     Group 2   3.5
3 B     Group 3   4  
4 C     Group 1   5  
5 C     Group 2   8.5
6 C     Group 3   1.5