handle missing observations for factor variables

missing_with_factors.R

if (!require("pacman")) install.packages("pacman")
#> Loading required package: pacman
pacman::p_load(tidyverse)

# con missing como NA -----------------------------------------------------

f1 <- tibble(variable = c("a", "a", NA, NA, "a", "b", 
                          NA, "c", "a", "c", "b"))
f1
#> # A tibble: 11 × 1
#>    variable
#>    <chr>   
#>  1 a       
#>  2 a       
#>  3 <NA>    
#>  4 <NA>    
#>  5 a       
#>  6 b       
#>  7 <NA>    
#>  8 c       
#>  9 a       
#> 10 c       
#> 11 b
f1 %>% count(variable)
#> # A tibble: 4 × 2
#>   variable     n
#>   <chr>    <int>
#> 1 a            4
#> 2 b            2
#> 3 c            2
#> 4 <NA>         3

f2 <- f1 %>% 
  mutate(variable = as.factor(variable)) 
f2
#> # A tibble: 11 × 1
#>    variable
#>    <fct>   
#>  1 a       
#>  2 a       
#>  3 <NA>    
#>  4 <NA>    
#>  5 a       
#>  6 b       
#>  7 <NA>    
#>  8 c       
#>  9 a       
#> 10 c       
#> 11 b
f2 %>% count(variable)
#> # A tibble: 4 × 2
#>   variable     n
#>   <fct>    <int>
#> 1 a            4
#> 2 b            2
#> 3 c            2
#> 4 <NA>         3
levels(f2$variable)
#> [1] "a" "b" "c"

# si no tengo los missing como NA -----------------------------------------

f1 <- tibble(variable = c("a", "a", NA, NA, "a", "b", 
                          NA, "missing", "a", "missing", "b"))
f1
#> # A tibble: 11 × 1
#>    variable
#>    <chr>   
#>  1 a       
#>  2 a       
#>  3 <NA>    
#>  4 <NA>    
#>  5 a       
#>  6 b       
#>  7 <NA>    
#>  8 missing 
#>  9 a       
#> 10 missing 
#> 11 b
f1 %>% count(variable)
#> # A tibble: 4 × 2
#>   variable     n
#>   <chr>    <int>
#> 1 a            4
#> 2 b            2
#> 3 missing      2
#> 4 <NA>         3

f2 <- f1 %>% 
  mutate(variable = as.factor(variable)) 
f2 %>% count(variable)
#> # A tibble: 4 × 2
#>   variable     n
#>   <fct>    <int>
#> 1 a            4
#> 2 b            2
#> 3 missing      2
#> 4 <NA>         3
levels(f2$variable)
#> [1] "a"       "b"       "missing"

pacman::p_load(naniar)
f3 <- f1 %>% 
  naniar::replace_with_na(replace = list(variable = "missing"))
f3
#> # A tibble: 11 × 1
#>    variable
#>    <chr>   
#>  1 a       
#>  2 a       
#>  3 <NA>    
#>  4 <NA>    
#>  5 a       
#>  6 b       
#>  7 <NA>    
#>  8 <NA>    
#>  9 a       
#> 10 <NA>    
#> 11 b

f4 <- f3 %>% 
  mutate(variable = as.factor(variable)) 
f4
#> # A tibble: 11 × 1
#>    variable
#>    <fct>   
#>  1 a       
#>  2 a       
#>  3 <NA>    
#>  4 <NA>    
#>  5 a       
#>  6 b       
#>  7 <NA>    
#>  8 <NA>    
#>  9 a       
#> 10 <NA>    
#> 11 b
f4 %>% count(variable)
#> # A tibble: 3 × 2
#>   variable     n
#>   <fct>    <int>
#> 1 a            4
#> 2 b            2
#> 3 <NA>         5
levels(f4$variable)
#> [1] "a" "b"

# hacer missing explicito -------------------------------------------------

f2
#> # A tibble: 11 × 1
#>    variable
#>    <fct>   
#>  1 a       
#>  2 a       
#>  3 <NA>    
#>  4 <NA>    
#>  5 a       
#>  6 b       
#>  7 <NA>    
#>  8 missing 
#>  9 a       
#> 10 missing 
#> 11 b
f3 <- f2 %>% 
  mutate(variable = fct_explicit_na(variable,
                                    na_level = "missing"))
f3
#> # A tibble: 11 × 1
#>    variable
#>    <fct>   
#>  1 a       
#>  2 a       
#>  3 missing 
#>  4 missing 
#>  5 a       
#>  6 b       
#>  7 missing 
#>  8 missing 
#>  9 a       
#> 10 missing 
#> 11 b
f3 %>% count(variable)
#> # A tibble: 3 × 2
#>   variable     n
#>   <fct>    <int>
#> 1 a            4
#> 2 b            2
#> 3 missing      5
levels(f3$variable)
#> [1] "a"       "b"       "missing"