Replace NA values with median by group

Question

I have used the below tapply function to get the median of Age based on Pclass.

Now how can I impute those median values to NA values based on Pclass?

tapply(titan_train$Age, titan_train$Pclass, median, na.rm=T)

Desired output

Please give a [minimal reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example). — markus, Dec 01 '18 at 12:25

score 3 · Accepted Answer · answered Dec 01 '18 at 13:37

3

Here is another base R approach that uses replace and ave.

df1 <- transform(df1,
                 Age = ave(Age, Pclass, FUN = function(x) replace(x, is.na(x), median(x, na.rm = T))))
df1
#   Pclass Age
# 1      A   1
# 2      A   2
# 3      A   3
# 4      B   4
# 5      B   5
# 6      B   6
# 7      C   7
# 8      C   8
# 9      C   9

Same idea but using data.table

library(data.table)
setDT(df1)
df1[, Age := as.integer(replace(Age, is.na(Age), median(Age, na.rm = T))), by = Pclass]
df1

data

df1 <- data.frame(Pclass = rep(LETTERS[1:3], each = 3),
                  Age = 1:9)
df1$Age[c(FALSE, TRUE, FALSE)] <- NA
df1
#   Pclass Age
# 1      A   1
# 2      A  NA
# 3      A   3
# 4      B   4
# 5      B  NA
# 6      B   6
# 7      C   7
# 8      C  NA
# 9      C   9

answered Dec 01 '18 at 13:37

markus

25,843
5
39
58

1

Thanks a lot!!! Tried data.table code and it worked perfectly fine once I removed as.integer. – Suhas U Dec 01 '18 at 14:12
You're welcome. Rejected your edit such that the example I gave is self-contained. You'd get an error if'd run it without `as.integer`. Best – markus Dec 01 '18 at 19:02
Ok, not a problem! – Suhas U Dec 02 '18 at 03:58

score 0 · Answer 2 · answered Dec 01 '18 at 13:07

Try the following.

set.seed(1)
df1 <- data.frame(Pclass = sample(1:3, 20, TRUE),
                  Age = sample(c(NA, 20:40), 20, TRUE, prob = c(10, rep(1, 21))))

new <- ave(df1$Age, df1$Pclass, FUN = function(x) median(x, na.rm = TRUE))
df1$Age[is.na(df1$Age)] <- new[is.na(df1$Age)]

Final clean up.

rm(new)

Replace NA values with median by group

2 Answers2

Linked

Related