dplyr: group_by & mutate variable can't call mean/sd functions

Question

Sorry if the title is a bit convoluted, as I didn't know how else to explain this issue. Basically, I'm attempting to mutate a percent of group variable utilizing dplyr. However, I'm running into an issue where the new calculated variable appears numeric, and even calculates when using summary(), but will not allow me to call mean() or sd() without throwing me the following error:

Warning message:
In mean.default(., group_pct) :
  argument is not numeric or logical: returning NA

Here are some examples of what is going on.

data(mtcars)

mtcars %>% 
  group_by(cyl) %>% 
  mutate(group_pct = hp / sum(hp)) %>% 
  summary()

Note: group_pct is calculating correctly when called via summary()...

data(mtcars)

mtcars %>% 
  group_by(cyl) %>% 
  mutate(group_pct = hp / sum(hp)) %>% 
  mean(group_pct)

...but when I call for the mean here, it cannot complete the function. Even when I use ungroup() and/or na.rm = TRUE, the function still doesn't work. I don't understand what the issue is here.

EDIT: For clarification, I'm hoping to do something like this...

mtcars %>% 
  group_by(cyl) %>% 
  mutate(group_pct = hp / sum(hp)) %>% 
  paste0('Words: ', mean(group_pct))

Hoping for this final result:

Words: 0.09375

...which I don't think I can use summarize() for, hence my non-inclusion of it from the start. Apologies for any inconveniences.

From my personal experience, expected output is most helpful. Do you just want a character vector that is the length of the number of rows of `mtcars` with each element of the form `"Words: #.###"`? — zack, Sep 12 '18 at 15:10
My bad, edited again for expected output. I'm hoping for one phrase and one numerical result which is the average of the rowwise group proportions. — medavis6, Sep 12 '18 at 15:21

zack · Answer 1 · 2018-09-12T16:21:21.457

4

Per OP's Clarification:

mtcars %>% 
  group_by(cyl) %>% 
  mutate(group_pct = hp / sum(hp)) %>%
  pull(group_pct) %>%
  mean() %>%
  paste0("Words: ", .)

[1] "Words: 0.09375"

edited Sep 12 '18 at 16:21

answered Sep 12 '18 at 14:51

zack

5,205
1
19
25

score 3 · Accepted Answer · answered Sep 12 '18 at 15:32

3

You want the baseR function with().

mtcars %>% 
  group_by(cyl) %>% 
  mutate(group_pct = hp / sum(hp)) %>%
  with(paste0('Words: ', mean(group_pct)))

[1] "Words: 0.09375"

The issue with your original attempt is that group_pct is not defined in the global environment, so you get the error message, when it can't locate it in the lookup.

with is the syntactic sugar that tells R to evaluate the paste0() function call within the environment of the data frame being passed by pipe. So it finds group_pct and returns your expected result.

answered Sep 12 '18 at 15:32

Nate

10,361
3
33
40

You really understood OP! – nghauran Sep 12 '18 at 15:40
It's a warning message, not an error, since they are passing group_pct as the second arg to mean, like `mean(mtcars, absolutely + any + gibberish)` or `mean(mtcars)`. Also, maybe of interest: with magrittr you can use a different pipe operator instead of with: `mtcars %$% mean(disp)` – Frank Sep 12 '18 at 16:14

divibisan · Answer 3 · 2018-09-12T16:05:15.430

The problem is with how you're piping into the mean function. Look at your error message:

Warning message:
In mean.default(., group_pct)

You're trying to get the mean of the group_pct column of the piped data frame, but instead what's happening is mean is receiving both the entire piped data frame (the . represents the output of the pipe) and an object group_pct which does may or may not exist.

Take a look at this answer: https://stackoverflow.com/a/38475455/8366499

If you want to subset the piped data.frame in the mean function, you need to use curly braces {} so the call to mean gets treated like an expression, not a function call. Then, you can subset the . object as desired:

mtcars %>% 
    group_by(cyl) %>% 
    mutate(group_pct = hp / sum(hp)) %>% 
    {mean(.$group_pct)} %>%
    paste0('Words: ', .)

[1] "Words: 0.09375"

Sal-laS · Answer 4 · 2018-09-12T16:57:58.393

0

library(tidyverse)
library(purrr)

mtcars %>% 
  mutate(group_pct = hp / sum(hp)) %>% 
  summarise_all(mean) %>%
  select(group_pct) %>%
  map(function(x) paste0(" Word ", x))

and the result is:

"Word 0.03125"

edited Sep 12 '18 at 16:57

answered Sep 12 '18 at 14:48

Sal-laS

11,016
25
99
169

1

@medavis6 maybe it is a little lat, but now my answer works as well :( – Sal-laS Sep 12 '18 at 16:15

dplyr: group_by & mutate variable can't call mean/sd functions

4 Answers4

Per OP's Clarification: