How do I change NA into Column median

Question

A data frame contains 123 columns, and each columns have at least 1 NA value.

I want these NA values to be raplaced into column median. because there are so many columns, i cannot write a code using each column name.

so i tried to use 'apply' to solve this but it didn't work.

data2[-1]<-lapply(data2[-1],function(x)x - median(x,na.rm=TRUE))

it says it doesn't work since it is data frame, not numeric.

Please provide example data as plain text, not images, so users can copy/paste. You can paste the output from `dput(data2)` or if large, `dput(head(data2))`. — neilfws, Jun 05 '17 at 05:41
**This is not a duplicate, so don't VtC.** @RonakShah: the question and answer you cite is about matrices not dataframes, although the asker had meant to ask about about dataframes, but they didn't. I retitled it *"Replacing NA's in each column of matrix with the median of that column"*. So we do actually do need a canonical example for a dataframe. Might as well use this question. — smci, Jun 05 '17 at 06:05

akrun · Answer 1 · 2017-06-05T06:04:36.997

1

We can use na.aggregate

 library(zoo)
 j1 <- sapply(df1, is.numeric)
 df1[j1] <- na.aggregate(df1[j1], FUN = median)

edited Jun 05 '17 at 06:04

answered Jun 05 '17 at 05:55

akrun

874,273
37
540
662

well, that didn't work. – Equinox Jun 05 '17 at 05:59
@Equinox can u be specific instead of `that didn't work` – akrun Jun 05 '17 at 06:00
console says : In mean.default(x[!is.na(x)]) : argument is not numeric or logical: returning NA – Equinox Jun 05 '17 at 06:03
1

@akrun I expect it's the usual have to assign the output `df1 <- na.aggregate(...)` – smci Jun 05 '17 at 06:03
@Equinox That is because you are applying for the whole data. I assumed that you had only numeric columns. In this case, apply it to selected columns – akrun Jun 05 '17 at 06:04
@Equinox Try it now – akrun Jun 05 '17 at 06:04
@akrun k, i ll try it. – Equinox Jun 05 '17 at 06:09
@Equinox I tried it on your dput example, it is working for me – akrun Jun 05 '17 at 06:24
@akrun no error found, but NA value didn't change – Equinox Jun 05 '17 at 07:40
@Equinox Then, please check whether the `NA` is real NA or `"NA"` – akrun Jun 05 '17 at 07:47
1

@akrun yepp, you got it. Checking with is.na, all NA is false. I guess i should change "NA" into real NA – Equinox Jun 05 '17 at 08:20

Sahban Tariq · Answer 2 · 2017-06-05T07:15:15.177

1

We can use map2_df

library(purrr)
df <- data.frame(a = c(1, 2, 3), b = c(2, NA, 9), c = c(NA, 3, 5), d = c(0, 4, NA))
purrr::map2_df(df, purrr::dmap(df, median, na.rm = TRUE), function(x, y) ifelse(is.na(x), y, x))

edited Jun 05 '17 at 07:15

answered Jun 05 '17 at 06:26

Sahban Tariq

35
7

AK88 · Accepted Answer · 2017-06-05T06:19:04.197

0

for(i in 1:ncol(df)){
  df[is.na(df[,i]), i] <- median(df[,i], na.rm = TRUE)
}

edited Jun 05 '17 at 06:19

answered Jun 05 '17 at 05:55

AK88

2,946
2
12
31

i have 123 columns i can't do $ function on my code that is too unefficient – Equinox Jun 05 '17 at 05:58
does my update work? – AK88 Jun 05 '17 at 06:06
well it made another error, but at least using 'for' like C gives me some idea. thx for that. – Equinox Jun 05 '17 at 06:12
what error did you get? – AK88 Jun 05 '17 at 06:14
always welcome )) – AK88 Jun 06 '17 at 10:06
May I ask something? – Equinox Jun 06 '17 at 11:29
of course,what's up? – AK88 Jun 06 '17 at 11:31
I put the same code for my new data.frame that contains instead of NA and this code stops working. it says they needs digital data. – Equinox Jun 06 '17 at 11:32
should I fix the data.frame so that it gets NA than or should i fix the code you wrote? – Equinox Jun 06 '17 at 11:33
I think you'd have to include some lines to get it work. Maybe fix `` to `NA`? You can use `gsub()` or other options. – AK88 Jun 06 '17 at 11:34

How do I change NA into Column median

3 Answers3

Linked