Suggestion to replace NA with the mode

Question

In prior experiments, I occasionally substituted the mean for NA. This code has worked so far.

df <- transform(df,
                 var_1 = ave(var_1, group,
                             FUN = function(x) replace(x, is.na(x),
                                                       mean(x, na.rm = T))))
df

My data is also categorized.

df <- data.frame(group = c(rep("cake", 5), rep("cookie",5)),
                    var_1 = c(1, 2, NA, 3, 1, 8, 9, NA, 7, 8))
df

The result I'm looking for.

    group var_1
1    cake     1
2    cake     2
3    cake     1
4    cake     3
5    cake     1
6  cookie     8
7  cookie     9
8  cookie     8
9  cookie     7
10 cookie     8

I tried using the dplyr package to replace NA with the mode. However, it did not work. Instead, I received an error message.

# not working
library(dplyr)
df %>% group_by(group) %>% mutate(var_1 = na.aggregate(var_1, FUN = mode))

Also, this code isn't working either.

# not working
library(dplyr)
df %>% group_by(group) %>% mutate(var_1 = if_else(is.na(var_1), 
                         calc_mode(var_1), var_1))

Here's a sample of an error.

Error: Problem with `mutate()` column `var_1`.
ℹ `var_1 = if_else(is.na(var_1), calc_mode(var_1), var_1)`.
x could not find function "calc_mode"
ℹ The error occurred in group 1: group = "cake".

Error: Problem with `mutate()` column `var_1`.
ℹ `var_1 = na.aggregate(var_1, FUN = mode)`.
x could not find function "na.aggregate"
ℹ The error occurred in group 1: group = "cake".

Any ideas would be greatly appreciated. Thank you very much for your advice.

Both of those errors say you're using functions that aren't loaded. Do you still get an error when you load the packages you're trying to work with? — camille, Dec 18 '21 at 01:01
Yes. I am certain that I correctly loaded the package. I greatly thank you for your suggestion. — cccn, Dec 18 '21 at 05:03

score 1 · Accepted Answer · answered Dec 18 '21 at 01:03

na_replace_Mode <- function(x) {
  ux <- unique(na.omit(x))
  x[is.na(x)] <- ux[which.max(tabulate(match(x, ux)))]
  x
}

transform(df, var_1 = ave(var_1, group, FUN = na_replace_Mode))

   group var_1
1    cake     1
2    cake     2
3    cake     1
4    cake     3
5    cake     1
6  cookie     8
7  cookie     9
8  cookie     8
9  cookie     7
10 cookie     8

You could also do:

Mode <- function(x) {
  x <- na.omit(x)
  ux <- unique(x)
  ux[which.max(tabulate(match(x, ux)))]
}

df %>%
   group_by(group) %>%
   mutate(var_1 = replace_na(Mode(var_1)))
# A tibble: 10 x 2
# Groups:   group [2]
   group  var_1
   <chr>  <dbl>
 1 cake       1
 2 cake       2
 3 cake       1
 4 cake       3
 5 cake       1
 6 cookie     8
 7 cookie     9
 8 cookie     8
 9 cookie     7
10 cookie     8

Suggestion to replace NA with the mode

1 Answers1