CODEWITHSUNDEEP
rcounttidyverse
GaB
GaB

Reputation: 1132

why I do not get counts over two numerical columns grouped by other categorical vars with tidyverse only?

I have tried to to counts of two numerical variables and did not succeed. Without this I cannot achieve the Percentages, which I hope I am able to get with your help. I am trying to do this only with tidyverse.

This is the error I got, with the code provided:

 test_sum <- test_data_3 %>%
    dplyr::group_by(across(where(is.factor))) %>% 
    dplyr::summarise(across(where(is.numeric())))


Error: Problem with `summarise()` input `..1`.
ℹ `..1 = across(where(is.numeric()))`.
x 0 arguments passed to 'is.numeric' which requires 1
Run `rlang::last_error()` to see where the error occurred.

The I tried another code:

test_sum <- test_data_3 %>%
    dplyr::group_by(provider_name, type, st_nst) %>% 
    dplyr::summarise(across(where(is.numeric())))

Error: Problem with `summarise()` input `..1`.
ℹ `..1 = across(where(is.numeric()))`.
x 0 arguments passed to 'is.numeric' which requires 1
ℹ The error occurred in group 1: provider_name = "BLACKB", type = "stri", st_nst = "NST".

And this is the stack overflow source I have inspired to try the previous codes: Group by multiple columns and sum other multiple columns

And this is the type of data I have:

dput(test_data_3)
structure(list(financial_year = c(1920, 1920, 1920, 1920, 1920, 
1920, 1920, 1920, 1920, 1920, 1920, 1920, 1920, 1920, 1920, 1920, 
1920, 1920, 1920, 1920), provider_name = c("LIVEW", "MANCHE", 
"MANCHE", "MANCHE", "MANCHE", "MANCHE", "MANCHE", "MANCHE", "MANCHE", 
"SOUTH", "LANCA", "COUNTY", "BUCKINGT", "BLACKB", "BURNLEY", 
"ROYAL", "THE", "LOUTH", "IMPERIAL", "WESTERN"), type = c("non_stringent", 
"non_stringent", "non_stringent", "non_stringent", "non_stringent", 
"non_stringent", "non_stringent", "non_stringent", "non_stringent", 
"non_stringent", "stri", "stri", "stri", "stri", "stri", "stri", 
"stri", "stri", "stri", "stri"), eld = c(0, 326, 343, 43, 61, 
46, 1, 3, 3, 1, 313, 671, 329, 389, 3, 376, 306, 0, 409, 589), 
    ed = c(1, 23, 23, 0, 2, 0, 1, 0, 0, 0, 7, 3, 4, 4, 0, 0, 
    2, 1, 3, 1), st_nst = c("ST", "STI", "ST", "ST", "ST", "ST", 
    "ST", "ST", "ST", "ST", "NST", "NST", "NSt", "NST", "NST", 
    "NST", "NST", "NST", "NST", "NST")), class = c("spec_tbl_df", 
"tbl_df", "tbl", "data.frame"), row.names = c(NA, -20L), spec = structure(list(
    cols = list(financial_year = structure(list(), class = c("collector_double", 
    "collector")), trust_code = structure(list(), class = c("collector_character", 
    "collector")), provider_name = structure(list(), class = c("collector_character", 
    "collector")), prim_diag = structure(list(), class = c("collector_character", 
    "collector")), type = structure(list(), class = c("collector_character", 
    "collector")), elective_discharge = structure(list(), class = c("collector_double", 
    "collector")), emergency_admission = structure(list(), class = c("collector_double", 
    "collector")), st_nst = structure(list(), class = c("collector_character", 
    "collector"))), default = structure(list(), class = c("collector_guess", 
    "collector")), skip = 1L), class = "col_spec"))

Or another way of visualising is as such:

test_data_3
# A tibble: 20 x 6
   financial_year provider_name type            eld    ed st_nst
            <dbl> <chr>         <chr>         <dbl> <dbl> <chr> 
 1           1920 LIVEW         non_stringent     0     1 ST    
 2           1920 MANCHE        non_stringent   326    23 STI   
 3           1920 MANCHE        non_stringent   343    23 ST    
 4           1920 MANCHE        non_stringent    43     0 ST    
 5           1920 MANCHE        non_stringent    61     2 ST    
 6           1920 MANCHE        non_stringent    46     0 ST    
 7           1920 MANCHE        non_stringent     1     1 ST    
 8           1920 MANCHE        non_stringent     3     0 ST    
 9           1920 MANCHE        non_stringent     3     0 ST    
10           1920 SOUTH         non_stringent     1     0 ST    
11           1920 LANCA         stri            313     7 NST   
12           1920 COUNTY        stri            671     3 NST   
13           1920 BUCKINGT      stri            329     4 NSt   
14           1920 BLACKB        stri            389     4 NST   
15           1920 BURNLEY       stri              3     0 NST   
16           1920 ROYAL         stri            376     0 NST   
17           1920 THE           stri            306     2 NST   
18           1920 LOUTH         stri              0     1 NST   
19           1920 IMPERIAL      stri            409     3 NST   
20           1920 WESTERN       stri            589     1 NST

Can someone explain the mistakes I do? Is there a way to achieve the counts first and then the percentages of the 2 numerical columns, namely eld & ed grouped by provider_name, type, st_nst. I mean these two columns to be added together into a new one based on the group by variables.

Upvotes: 1

Views: 46

Answers (1)

akrun
akrun

Reputation: 887213

There was no function passed into across. If the intention is to select the columns

library(dplyr)
test_data_3 %>%
    dplyr::group_by(across(where(is.factor))) %>% 
    dplyr::select(where(is.numeric))

Suppose, we want to get the sum of those numeric columns

test_data_3 %>%
    dplyr::group_by(across(where(is.factor))) %>% 
    dplyr::summarise(across(where(is.numeric), sum))

Update

If we want to get the sum per row of numeric columns, select the numeric columns (where(is.numeric)) from the data (cur_data() - would be more correct as it can also work when there are group attributes or use .), get the row wise sum with rowSums

test_data_3 %>% 
      mutate(count = select(cur_data(), where(is.numeric)) %>% 
                  rowSums(na.rm = TRUE))

Upvotes: 1

Related Questions