Part 8 Handling missing values

  • drop_na: drop rows containing missing values.

Create a tibble that contains missing (NA) values:

dfwithNA <- tibble(x = c(1, 2, NA, 5), 
                   y = c("a", NA, "b", "c"))

Remove rows that contain NA values with drop_na:

drop_na(dfwithNA)
## # A tibble: 2 x 2
##       x y    
##   <dbl> <chr>
## 1     1 a    
## 2     5 c

  • replace_na: change NA values to a selected value (one per column):
# replace NAs by 0s in column "x", and by "k" in column "y"
replace_na(dfwithNA, 
           list(x=0, y="k"))
## # A tibble: 4 x 2
##       x y    
##   <dbl> <chr>
## 1     1 a    
## 2     2 k    
## 3     0 b    
## 4     5 c

  • complete: turns implicit missing values into explicit missing values:
df <- tibble(
  patient = c("Patient1", "Patient1", "Patient2", "Patient3", "Patient3"),
  treatment = c("A", "B", "A", "A", "B"),
  value1 = 1:5,
  value2 = 4:8
)

Here we are missing a row for Patient2 / Treatment B: add it and fill in with NA values:

complete(df, 
         patient, treatment) # columns to expand
## # A tibble: 6 x 4
##   patient  treatment value1 value2
##   <chr>    <chr>      <int>  <int>
## 1 Patient1 A              1      4
## 2 Patient1 B              2      5
## 3 Patient2 A              3      6
## 4 Patient2 B             NA     NA
## 5 Patient3 A              4      7
## 6 Patient3 B              5      8

If you want implicit missing values to be filled by something else than NA, use the fill parameter:

# we will fill in missing values with 0s in column "value1", and with NAs in column "value2"
complete(df, 
         patient, treatment, 
         fill=list(value1=0, value2=NA))
## # A tibble: 6 x 4
##   patient  treatment value1 value2
##   <chr>    <chr>      <dbl>  <int>
## 1 Patient1 A              1      4
## 2 Patient1 B              2      5
## 3 Patient2 A              3      6
## 4 Patient2 B              0     NA
## 5 Patient3 A              4      7
## 6 Patient3 B              5      8

  • In practice: what if you have NA values, along with empty cells and “customized” missing values?
dfwithNA2 <- tibble(col1=c(1, 2, NA, 5, "", 4), 
                    col2=c("a", NA, "b", "c", "d", "missing"))

Replace empty cells and “customized” missing values with NA with na_if:

dfwithNA2 %>% 
   mutate(col1=na_if(col1, ""), col2=na_if(col2, "missing"))
## # A tibble: 6 x 2
##   col1  col2 
##   <chr> <chr>
## 1 1     a    
## 2 2     <NA> 
## 3 <NA>  b    
## 4 5     c    
## 5 <NA>  d    
## 6 4     <NA>

HANDS-ON

Back to the starwars data set:

  • Replace NA in hair_color with “unknown.”
  • Remove rows that still contain NA values.
Answer
# Replace NA in `hair_color` with "unknown".
# Remove rows that still contain NA values.
replace_na(starwars, list(hair_color="unknown")) %>%
  drop_na()