Find relative frequencies of summarized columns in R-CodePudding

I need to get the relative frequencies of a summarized column in R. I've used dplyr's summarize to find the total of each grouped row, like this:

data %>%
  group_by(x) %>%
  summarise(total = sum(dollars))

     x                    total 
   <chr>                 <dbl>
 1 expense 1              3600 
 2 expense 2              2150 
 3 expense 3              2000

But now I need to create a new column for the relative frequencies of each total row to get this result:

     x                   total     p
   <chr>                 <dbl>   <dbl>
 1 expense 1              3600   46.45%
 2 expense 2              2150   27.74%
 3 expense 3              2000   25.81%

I've tried this:

data %>%
  group_by(x) %>%
  summarise(total = sum(dollars), p = scales::percent(total/sum(total))

and this:

data %>%
  group_by(x) %>%
  summarise(total = sum(dollars), p = total/sum(total)*100)

but the result is always this:

     x                   total     p
   <chr>                 <dbl>   <dbl>
 1 expense 1              3600    100%
 2 expense 2              2150    100%
 3 expense 3              2000    100%

The problem seems to be the summarized total column that may be affecting the results. Any ideas to help me? Thanks

CodePudding user response：

You get 100% because of the grouping. However, after you've summarized, dplyr will drop the one level of grouping. Meaning that if you e.g. do mutate() after, you get the results you need:

library(dplyr)

data <- tibble(
  x = c("expense 1", "expense 2", "expense 3"),
  dollars = c(3600L, 2150L, 2000L)
)


data %>%
  group_by(x) %>%
  summarise(total = sum(dollars)) %>% 
  mutate(p = total/sum(total)*100)


# A tibble: 3 x 3
  x         total     p
  <chr>     <int> <dbl>
1 expense 1  3600  46.5
2 expense 2  2150  27.7
3 expense 3  2000  25.8

CodePudding user response：

You get 100% because it calculates the total of that particular group. You need to ungroup. Assuming that you want to divide by total entries just divide by nrow(df).

data %>%
  group_by(x) %>%
  summarise(total = sum(dollars), p = total/nrow(data)*100)

CodePudding user response：

After the first sum, ungroup and create p with mutate.

iris %>%
  group_by(Species) %>%
  summarise(total = sum(Sepal.Length)) %>%
  ungroup() %>%
  mutate(p = total/sum(total)*100)
## A tibble: 3 x 3
#  Species    total     p
#  <fct>      <dbl> <dbl>
#1 setosa      250.  28.6
#2 versicolor  297.  33.9
#3 virginica   329.  37.6