R: Calculating Percentiles For Multiple Groups

Question

I am working with the R programming language.

I have the following dataset:

set.seed(123)

library(dplyr)
var1 = rnorm(10000, 100,100)
var2 = rnorm(10000, 100,100)
var3 = rnorm(10000, 100,100)
var4 = rnorm(10000, 100,100)
id = 1:10000

final = data.frame(id, var1, var2, var3, var4)

final = final %>%
  mutate(class1 = case_when(var1 < mean(var1) ~ "A",
                             TRUE ~ "B")) %>% 
mutate(class2 = case_when(var2 < mean(var2) ~ "C",
                             TRUE ~ "D"))

I want to calculate deciles for var3 and var4 based on every unique combination of class1 and class2.

As I understand, this means:

For all rows WHERE class1 = A AND class2 = C, calculate/assign deciles for var3 and var4
For all rows WHERE class1 = A AND class2 = D, calculate/assign deciles for var3 and var4
For all rows WHERE class1 = B AND class2 = C, calculate/assign deciles for var3 and var4
For all rows WHERE class1 = B AND class2 = D, calculate/assign deciles for var3 and var4

Here is the R code I wrote for this:

final = final %>%
group_by(class1, class2) %>%
  mutate(class3 = case_when(ntile(var3, 10) == 1 ~ "one",
                             ntile(var3, 10) == 2 ~ "two",
                             ntile(var3, 10) == 3 ~ "three",
                             ntile(var3, 10) == 4 ~ "four",
                             ntile(var3, 10) == 5 ~ "five",
                             ntile(var3, 10) == 6 ~ "six",
                             ntile(var3, 10) == 7 ~ "seven",
                             ntile(var3, 10) == 8 ~ "eight",
                             ntile(var3, 10) == 9 ~ "nine",
                             ntile(var3, 10) == 10 ~ "ten")) %>%
  mutate(class4 = case_when(ntile(var4, 10) == 1 ~ "one",
                             ntile(var4, 10) == 2 ~ "two",
                             ntile(var4, 10) == 3 ~ "three",
                             ntile(var4, 10) == 4 ~ "four",
                             ntile(var4, 10) == 5 ~ "five",
                             ntile(var4, 10) == 6 ~ "six",
                             ntile(var4, 10) == 7 ~ "seven",
                             ntile(var4, 10) == 8 ~ "eight",
                             ntile(var4, 10) == 9 ~ "nine",
                             ntile(var4, 10) == 10 ~ "ten"))

Can someone please tell me if I have done this correctly?

Thanks!

akrun · Accepted Answer · 2022-12-26T19:37:09.470

3

Instead of doing the case_when it can be done easily with english

library(dplyr)
library(stringr)
final %>%
   group_by(class1, class2) %>% 
   mutate(across(var3:var4, 
         ~ as.character(english::english(ntile(.x, 10))),
       .names = "{str_replace(.col, 'var', 'class')}")) %>% 
   ungroup

-output

# A tibble: 10,000 × 9
      id  var1  var2    var3  var4 class1 class2 class3 class4
   <int> <dbl> <dbl>   <dbl> <dbl> <chr>  <chr>  <chr>  <chr> 
 1     1  44.0 337.    16.4   80.6 A      D      three  five  
 2     2  77.0  83.3   77.9  126.  A      C      five   six   
 3     3 256.  193.  -110.    46.2 B      D      one    four  
 4     4 107.   43.2  -66.8  -17.9 B      C      one    two   
 5     5 113.  123.    -9.80 190.  B      D      two    nine  
 6     6 272.  213.   -66.6   98.4 B      D      one    six   
 7     7 146.  238.    95.0  118.  B      D      five   six   
 8     8 -26.5  76.7  256.   160.  A      C      ten    eight 
 9     9  31.3 -60.1   59.5  126.  A      C      four   six   
10    10  55.4  70.2  179.   130.  A      C      eight  seven 
# … with 9,990 more rows
# ℹ Use `print(n = ...)` to see more rows

edited Dec 26 '22 at 19:37

answered Dec 26 '22 at 19:34

akrun

874,273
37
540
662

1

@ akrun: thank you so much! but just to clarify - is my answer correct as well? – stats_noob Dec 26 '22 at 19:36
2

@ Tarjae: both yourself and akrun are masters for me! both of you are amazing R coders, even more amazing R teachers, and even more amazing humans that are selflessly teaching the world how to learn R! I can not thank both of you (and everyone else on this website) for all the contributions you have made! I deeply appreciate it! – stats_noob Dec 26 '22 at 19:38
1

@stats_noob As per your description, you are calculating the ntile for var3, var4, grouped by 'class3', 'class4'. But in your case, you are applying the `ntile` multiple times just to convert the value to 'one', 'two', etc – akrun Dec 26 '22 at 19:38
1

@Tarjae Sorry, I didn't see your comment. Can you show the link again – akrun Dec 26 '22 at 19:39
1

@ akrun: in the end, should my approach and your approach give you the same result? – stats_noob Dec 26 '22 at 19:47
1

@stats_noob I checked your method and mine `all.equal(final1, final2, check.attributes = FALSE)# [1] TRUE` – akrun Dec 26 '22 at 19:48
@ akrun: I posted a related question over here -if you have time, can you please take a look at it? Thank you so much! https://stackoverflow.com/questions/74959267/r-difficulty-calculating-percentiles – stats_noob Dec 31 '22 at 01:54
@stats_noob I posted a solution, but it is not clear where your error is occuring. Is it based on the number of observations in the group and skip it, then the `possibly` could work – akrun Dec 31 '22 at 05:05

R: Calculating Percentiles For Multiple Groups

1 Answers1