Apply function to subset of columns using purrr and dplyr

Question

I am new to the purrr package, but I like the little I know about it.

Using only tidyverse packages, I would like to be able to add a column that is the result of a function applied to a subset of columns in a dataset.

Here is some toy data. A series of columns of factors

df <- data.frame(a_1 = factor(rep(letters[1:3], times = 5)),
                 a_2 = factor(rep(letters[1:3], times = 5)),
                 a_3 = factor(rep(letters[1:3], times = 5)),
                 b_1 = factor(rep(letters[1:3], times = 5)),
                 b_2 = factor(rep(letters[1:3], times = 5)),
                 b_3 = factor(rep(letters[1:3], times = 5)))

df

# output
#  a_1 a_2 a_3 b_1 b_2 b_3
# 1    a   a   a   a   a   a
# 2    b   b   b   b   b   b
# 3    c   c   c   c   c   c
# 4    a   a   a   a   a   a
# 5    b   b   b   b   b   b
# 6    c   c   c   c   c   c
# 7    a   a   a   a   a   a
# 8    b   b   b   b   b   b
# 9    c   c   c   c   c   c
# 10   a   a   a   a   a   a
# 11   b   b   b   b   b   b
# 12   c   c   c   c   c   c
# 13   a   a   a   a   a   a
# 14   b   b   b   b   b   b
# 15   c   c   c   c   c   c

The following function, via purr::map_df and dplyr::select cycles through the columns of df that start with a_, converts them to numeric class, finds the mean of those columns, then multiplies by 3.

rowMeans(purrr::map_df(.x = df %>% dplyr::select(grep("a_", names(.))),
                       .f = function(x) x <- as.numeric(x))*3)

# output
# [1] 3 6 9 3 6 9 3 6 9 3 6 9 3 6 9

This is the correct output, but is a vector.

Using a tidyverse function how do I add the result of my function to the existing df dataset as a new column, instead of as a vector?

Something involving dplyr::mutate I assume, but I can't work it out.

user_3962914 · Answer

You could use pmap_dbl :

library(dplyr)
library(purrr)

df %>%  
   mutate(mean_vec = pmap_dbl(select(., starts_with('a_')), 
                         ~mean(as.numeric(c(...)) * 3)))


#   a_1 a_2 a_3 b_1 b_2 b_3 mean_vec
#1    1   1   1   a   a   a        3
#2    2   2   2   b   b   b        6
#3    3   3   3   c   c   c        9
#4    1   1   1   a   a   a        3
#5    2   2   2   b   b   b        6
#6    3   3   3   c   c   c        9
#7    1   1   1   a   a   a        3
#8    2   2   2   b   b   b        6
#9    3   3   3   c   c   c        9
#10   1   1   1   a   a   a        3
#11   2   2   2   b   b   b        6
#12   3   3   3   c   c   c        9
#13   1   1   1   a   a   a        3
#14   2   2   2   b   b   b        6
#15   3   3   3   c   c   c        9

Or another option :

df %>%
  mutate_at(vars(starts_with('a')), as.numeric) %>%
  mutate(mean_vec = rowMeans(select(., starts_with('a_')) * 3))

Apply function to subset of columns using purrr and dplyr

1 Answers