How do I get the median of multiple columns in R with conditions (according to another column)

I'm a beginner in R and I would like to know how to do the following task:

I want to replace the missing values of my dataset by the median for all the columns of my dataset. However, for each column, I want the median of a certain category (depending on another column).My dataset is as follows

structure(list(Country = structure(1:5, .Label = c("Afghanistan", 
"Albania", "Algeria", "Andorra", "Angola"), class = "factor"), 
    CountryID = 1:5, Continent = c(1L, 2L, 3L, 2L, 3L), Adolescent.fertility.rate.... = c(151L, 
    27L, 6L, NA, 146L), Adult.literacy.rate.... = c(28, 98.7, 
    69.9, NA, 67.4)), class = "data.frame", row.names = c(NA, 
-5L))

So for each of the columns, I want to replace the missing values by the median of the values in the specific continent.

Solution

We can use dplyr::mutate_at to replace NAs in each column (except Continent and the non numeric column Country) with the median for its Continent group

df <- structure(list(Country = structure(1:5, .Label = c("Afghanistan",  "Albania", "Algeria", "Andorra", "Angola"), class = "factor"), 
               CountryID = 1:5, Continent = c(1L, 2L, 3L, 2L, 3L),
               Adolescent.fertility.rate.... = c(151L, 27L, 6L, NA, 146L),
               Adult.literacy.rate.... = c(28, 98.7, 69.9, NA, 67.4)), class = "data.frame", row.names = c(NA, -5L))

library(dplyr)
df %>%
  group_by(Continent) %>% 
  mutate_at(vars(-group_cols(), -Country), ~ifelse(is.na(.), median(., na.rm = TRUE), .)) %>% 
  ungroup()

Returns:

  # A tibble: 5 x 5
    Country     CountryID Continent Adolescent.fertility.rate.... Adult.literacy.rate....
    <fct>           <int>     <int>                         <int>                   <dbl>
  1 Afghanistan         1         1                           151                    28  
  2 Albania             2         2                            27                    98.7
  3 Algeria             3         3                             6                    69.9
  4 Andorra             4         2                            27                    98.7
  5 Angola              5         3                           146                    67.4

Explanation: First we group the data.frame df by Continent. Then we mutate all columns except the grouping column (and Country which is not numeric) the following way: If is.na is TRUE, we replace it with the median, and since we are grouped, it's going to be the median for the Continent group (if its not NA we replace it with itself). Finally we call ungroup for good measure to get back a 'normal' tibble.