Calculating columns with row percentage after obtaining sums by group in dplyr

Question

Using dplyr I'm generating a simple summary table for two categories:

# Data
data("mtcars")
# Lib
require(dplyr)
# Summary
mt_sum <- mtcars %>%
  group_by(am, gear) %>%
  summarise(n = n()) %>%
  spread(key = am, value = n)

Which produces the desired results:

Source: local data frame [3 x 3]

   gear     0     1
  (dbl) (int) (int)
1     3    15    NA
2     4     4     8
3     5    NA     5

To the generated table I would like to add a set of columns that would have row percentages instead of the presently available totals.

Desired results

I would like for my table to look like that:

   gear     0     1   0per   1per
1     3    15    NA   100%   
2     4     4     8   33%    67%    
3     5    NA     5          100%

Attempts

I tried to achieve the following by adding the code:

mt_sum <- mtcars %>%
  group_by(am, gear) %>%
  summarise(n = n()) %>%
  spread(key = am, value = n) %>%
  mutate_each(funs(./rowSums(.)))

but it returns the following error:

Error: 'x' must be an array of at least two dimensions

Hence my question: how can I add extra columns with row percentage values in dplyr?

Side points

• I would prefer blank values instead of NAs
• The table could be easily build with use of CrossTable in gmodels but I would like to stay in dplyr as I want to keep as many transformations as possible in one place

Answer 1

I think this is what you need:

# Data
data("mtcars")
# Lib
require(dplyr)
require(tidyr)
require(scales) #for percent
# Summary
mtcars %>%
  group_by(am, gear) %>%
  summarise(n = n()) %>%
  spread(key = am, value = n) %>%
  #you need rowwise because this is a rowwise operation
  rowwise %>%
  #I find do to be the best function for ad-hoc things that 
  #have no specific dplyr function
  #I use do below to calculate the numeric percentages
  do(data.frame(.,
                per0 = .$`0` / sum(.$`0`, .$`1`, na.rm=TRUE),
                per1 = .$`1` / sum(.$`0`, .$`1`, na.rm=TRUE))) %>%
  #mutate here is used to convert NAs to blank and numbers to percentages
  mutate(per0 = ifelse(is.na(per0), '', percent(per0)),
         per1 = ifelse(is.na(per1), '', percent(per1)))

Output:

Source: local data frame [3 x 5]
Groups: <by row>

   gear    X0    X1  per0  per1
  (dbl) (int) (int) (chr) (chr)
1     3    15    NA  100%      
2     4     4     8 33.3% 66.7%
3     5    NA     5        100%

Answer 2

Here is a way to do it with reshaping:

library(dplyr) library(tidyr)

mtcars %>%
  count(gear, am) %>%
  mutate(percent = n / sum(n)) %>%
  gather(variable, value, 
         n, percent) %>%
  unite("new_variable", am, variable) %>%
  spread(new_variable, value)