use tabyl with group_by?

[randomgambit]

Hello @sfirke,

Thanks for this very nice package.

I am confused as to whether tabyl can be used with a group_by.

Here is the common usage: group_by some variable and print the output of tabyl(myvar) for each group.

I was surprisingly unable to do so with a naive group_by(mygroup) %>% tabyl(myvar)

Can I do that with tabyl? is it meant to be used with grouping?
Thanks!

[sfirke]

Hi! tabyl ungroups a grouped input, treating it as if it was ungrouped: #125

If you want to group by one variable and count another, you simply do a two-way tabyl:

> mtcars %>% tabyl(am, cyl)
 am 4 6  8
  0 3 4 12
  1 8 3  2

Then calculate your percentages etc. with adorn_ functions.

In terms of the numbers you get, I think you can produce the same #s from a 1-way tabyl in this 2-way tabyl. I'd say this is the "janitor-ish" solution.

But if you preferred the style of a list of 1-way tabyls, you could try:

> split(mtcars, mtcars$cyl) %>%
+   purrr::map(tabyl, am)
$`4`
 am n   percent
  0 3 0.2727273
  1 8 0.7272727

$`6`
 am n   percent
  0 4 0.5714286
  1 3 0.4285714

$`8`
 am  n   percent
  0 12 0.8571429
  1  2 0.1428571

I gather there are more tidyverse-oriented alternatives to split, say nesting from purrr or this proposed chop: https://coolbutuseless.bitbucket.io/2018/03/07/chop-a-tidyverse-style-split/

[randomgambit]

Hi, very interesting indeed. No I do now want to use a double sort here because I have quite a bit of different groups (so as you correctly guess I need to keep a long format here).

The purrr solution seems fine, although I wonder if janitor could/should handle that natively, by binding together the by-group dataframes shown in the purrr output