Noobie Noobie - 16 days ago 5
R Question

Dplyr: how to loop over specific columns whose names are in a list?

I have a dataframe that looks like this

set.seed(10)
sample <- data_frame(group = c('A','B','C','C',NA,'D'),
var_hello = rnorm(6),
var_how = rnorm(6),
var_are = rnorm(6),
var_you = rnorm(6),
var_buddy = rnorm(6))
# A tibble: 6 × 6
group var_hello var_how var_are var_you var_buddy
<chr> <dbl> <dbl> <dbl> <dbl> <dbl>
1 A 0.01874617 -1.2080762 -0.23823356 0.9255213 -1.2651980
2 B -0.18425254 -0.3636760 0.98744470 0.4829785 -0.3736616
3 C -1.37133055 -1.6266727 0.74139013 -0.5963106 -0.6875554
4 C -0.59916772 -0.2564784 0.08934727 -2.1852868 -0.8721588
5 <NA> 0.29454513 1.1017795 -0.95494386 -0.6748659 -0.1017610
6 D 0.38979430 0.7557815 -0.19515038 -2.1190612 -0.2537805


In my original dataset, there are many, many
var_something
variables.

I would like to
group_by('group')
and compute the
mean
of a subset of these
var_something
variables, but even this subset can be large. So I dont want to resort to typing manually each
mutate
for every variable.

In the example, I am interested in variables in the following list
['var_hello', 'var_are']


I dont know how to code that up efficiently in
dplyr
. In
Pandas
, one could simply write

for var in ['var_hello', 'var_are']:
sample[computation +'_' + var] = sample.groupby('group')[var].agg('mean')


Note how I can automatically create the new column names (of the form
computation_var_hello
) . What is the best way to achieve that in
dplyr
?

Many thanks!

Answer

You can do this simply by using group_by and summarize_each. You then specify which variables you want to summarize, then replace the prefix in the names using setNames.

sample %>%
   group_by(group) %>%
   summarize_each(funs(mean), var_hello, var_are) %>% 
   setNames(gsub("var_","computation_var_",colnames(.)))