Is there R-based codes to calculate median for groups in R-CodePudding

Here is a sample of my data:

dat<-read.table (text=" W1D1    W1D2    W1D3    MN1 MN2 MN3
11  13  12  A   A   B
11  14  17  B   B   B
12  17  10  C   C   B
10  10  19  C   C   B
19  12  18  D   C   B
19  18  10  A   D   C
18  10  17  B   D   D
16  16  18  C   D   D
19  17  20  C   D   A
16  14  17  D   D   A

", header=TRUE)

W1D1 corresponds to MN1, W1D2 corresponds to MN2 and W1D3 corresponds to MN3. I want to calculate the Median for the values of C and D. For example, MN1, has the values of 12,10,19,16,19,16 in W1D1. So I get a median of 16. So W1D1 goes to MN1, W1D2 goes to MN2 and W1D3 goes to MN3.
The outcome is :

   

 Median1    16
    Median2 15
    Median3 17

Is it possible to do it using R-based, if not what is the simplest option?

Considering I have 10 Medians but to simplify the matter I have provided 3 Medians

CodePudding user response：

I'm not sure I fully understood what you are looking for, but here is one attempt (although not very elegant).

df1 <- df  %>%
    pivot_longer(cols = starts_with("M")) %>%
    filter(value %in% c("C", "D")) %>%
    group_by(name) %>%
    summarise(across(starts_with("W1"), median)) %>%
    select(-name)
  
  
df2 <- tibble(median = df1 %>%
                  as.matrix() %>%
                  diag(),
                variable = names(df1))

-Result

> df2

# A tibble: 3 x 2
  median variable
   <dbl> <chr>   
1     16 W1D1    
2     15 W1D2    
3     17 W1D3

CodePudding user response：

Here is an approach using pmap_dfc and then pivoting the result.

library(tidyverse)

n <- ncol(dat) / 2

pmap_dfc(list(str_c("Median", 1:n), str_c('W1D', 1:n), str_c('MN', 1:n)),
     ~ filter(dat, .data[[..3]] %in% c('C', 'D')) %>% 
       summarise('{..1}' := median(.data[[..2]]))) %>%
  pivot_longer(everything())
#> # A tibble: 3 × 2
#>   name    value
#>   <chr>   <dbl>
#> 1 Median1    16
#> 2 Median2    15
#> 3 Median3    17

^{Created on 2021-12-30 by the reprex package (v2.0.1)}