CODEWITHSUNDEEP
rtidyverseforcats
user25494
user25494

Reputation: 1371

How to replace all <NA> values in a data.frame using forcats::fct_explicit_na()?

I have a data frame with 19 variables, 17 of which are factors. Some of these factors contain missing values, coded as NA. I would like to recode missings as a separate factor level "to_impute" using forcats::fct_explicit_na() for all factors in the data frame.

A small example with two factor variables:

df <- structure(list(loc_len = structure(c(NA, NA, NA, NA, NA, NA, 
1L, 1L, 3L, 1L), .Label = c("No", "< 5 sec", "5 sec - < 1 min", 
"1 - 5 min", "> 5 min", "Unknown duration"), class = "factor"), 
    AMS = structure(c(1L, 2L, NA, 1L, 1L, NA, NA, NA, NA, NA), .Label = c("No", 
    "Yes"), class = "factor")), .Names = c("loc_len", "AMS"), row.names = c(NA, 
-10L), class = c("tbl_df", "tbl", "data.frame"))

table(df$loc_len, useNA = "always")

              No          < 5 sec  5 sec - < 1 min        1 - 5 min          > 5 min Unknown duration             <NA> 
               3                0                1                0                0                0                6

The code below does this for two variables. I'd like to do this for all factor variables 'f_names' in the data frame. Is there a way to 'vectorize' fct_explicit_na()?

f_names <- names(Filter(is.factor, df))

 f_names
[1] "loc_len" "AMS"

The code below does what I want, but separately for each factor:

df_new <- df  %>% 
                    mutate(loc_len = fct_explicit_na(loc_len, na_level = "to_impute")) %>% 
                    mutate(AMS = fct_explicit_na(AMS, na_level = "to_impute"))

I'd like tables of this sort for all factors in the dataset, names in 'f_names' :

lapply(df_new, function(x) data.frame(table(x, useNA = "always")))

Now is:

$loc_len
                 x Freq
1               No    3
2          < 5 sec    0
3  5 sec - < 1 min    1
4        1 - 5 min    0
5          > 5 min    0
6 Unknown duration    0
7        to_impute    6
8             <NA>    0

$AMS
          x Freq
1        No    3
2       Yes    1
3 to_impute    6
4      <NA>    0

Upvotes: 2

Views: 3194

Answers (2)

user25494
user25494

Reputation: 1371

Even better, the elegant and idiomatic solution provided by:

https://github.com/tidyverse/forcats/issues/122

library(dplyr)
df = df %>% mutate_if(is.factor,
                      fct_explicit_na,
                      na_level = "to_impute")

Upvotes: 4

user25494
user25494

Reputation: 1371

After some trial and error, the code below does what I want.

library(tidyverse)

df[, f_names] <- lapply(df[, f_names], function(x) fct_explicit_na(x, na_level = "to_impute")) %>% as.data.frame

Upvotes: 0

Related Questions